/** * Mipham Code — Slash Command Registry * * All commands mirror Claude Code's UX so users need zero re-learning. * Commands marked [stub] are recognized but indicate WIP status. */ import type { QueryEngine } from '../core/engine' import type { MiphamConfig } from '../shared/index.ts' import { formatContextWindow } from '../shared/format' import type { SkillsLoader } from '../skills/loader' import { loadSkillUsage } from '../skills/usage' import type { PluginManager } from '../plugin/plugin-manager' import type { Message } from '../shared/types.js' import type { UpdateStatus } from '../shared/update' import { McpClient } from '../mcp/client' import { unregisterMcpServerTools } from '../mcp/registry' import { buildCapabilityReport } from '../core/capability-inventory' import { InstructionsLoader } from '../core/instructions' import { findDerivableSections, DERIVABLE_HINTS } from '../core/claude-md-audit' import { miphamHome, workflowScriptDir, workflowScriptDirs, worktreeRoot } from '../core/paths.ts' import { fixDoctor, fixConfig, fixCache, selectRepoClaudeFiles } from '../core/fix' import { fixCodeTarget } from '../core/fix-code' import { homedir } from 'node:os' import { runCrsiModification, approvePending, rejectPending, hasPending } from '../core/crsi-modify' import { produceCrsiProposal, produceRuleProposal, produceProseProposal, produceCrossoverProposal, selectCrsiSignal, collectSkillFiles, proseProposalId, hasProposedProse, appendProseProposal, clearProseProposals, LESSONS_FILE, MANAGED_RULES_FILE, } from '../core/crsi-producer' import { prefilterProposal } from '../core/proposal-guard' import { runEval, appendEvalScore, getContractHistory, diffContractHistory, renderContractDiff, } from '../core/eval-harness' import { listRewardFns } from '../core/reward-fn' import { runTaskPerformance, measureSkillDeltaRepeated, stripCodeFences, } from '../core/task-performance' import { randomUUID } from 'node:crypto' import { buildImprovementReport, formatCostLine, appendImprovement, readImprovements, improvementRate, predictionHitRate, setPendingVerdict, getPendingVerdict, shouldBlockApproval, } from '../core/improvement-track' import { NPM_UPDATE_COMMAND, PACKAGE_VERSION, COAUTHOR_TRAILER } from '../shared/index.ts' import { getPreference } from '../config/preferences' import { loadCrossSessionConfig, tryRestoreFromBackup } from '../config/loader' import { getWorkspaceTrust } from '../core/workspace-trust' import { getMemoryManager } from '../core/memory/memory-loader' import { resolveTelemetry, readTelemetrySettings, setTelemetryEnabled, setTelemetryEndpoint, resetInstallId, } from '../telemetry/consent' import { getTelemetryConsent, enableTelemetryNow } from '../telemetry/index' import { NO_ENDPOINT } from '../telemetry/endpoint' import { stripIndent } from './strip-indent.js' import { createT } from '../i18n-core/t' import type { TranslationMap } from '../i18n-core/types' import enUS from '../i18n-core/locales/en-US.json' with { type: 'json' } import zhCN from '../i18n-core/locales/zh-CN.json' with { type: 'json' } // Default t() fallback for plain-function contexts (tests, bootstrap). // When called from the React tree, ctx.t is populated from useI18n(). const defaultT = createT(enUS as unknown as TranslationMap, zhCN as unknown as TranslationMap) /** Resolve the best available t() function: prefer ctx.t (locale-aware from React), * fall back to module-level defaultT (en-US). */ function resolveT(ctx: CommandContext): (key: string, params?: Record) => string { return ctx.t && typeof ctx.t === 'function' ? ctx.t : defaultT } import { initCmd, setupCmd, recommendCmd, permissionsCmd, addDirCmd, securityCmd, promptAuditCmd, trustCmd, } from '../commands/project.js' import { themeCmd, releaseNotesCmd, ideCmd, terminalSetupCmd } from '../commands/environment.js' import { commitCmd, pushCmd, prCmd, issueCmd } from '../commands/git.js' import { suggestDirectories } from '../commands/cd-suggest.js' import { keysCmd } from '../commands/keys' import { workflowViewCmd, workflowWatchCmd } from '../commands/workflow-view.js' import { listActiveAutoloops, formatLoopRows } from '../commands/autoloop-journal.js' import { execSync } from 'node:child_process' import { MIPHAM_DIR, OLLAMA_PRESET_MODELS } from '../shared/constants' import { renameActiveSession } from '../agent/cross-session/discovery' export interface CommandContext { engine: QueryEngine config: MiphamConfig providerId: string modelId: string version: string sessionId: string // Callbacks for commands that mutate App state setSessionTitle: (title: string) => void setFastMode: (on: boolean) => void setEffort: (level: string) => void setFocusMode: (on: boolean) => void setGoal: (text: string) => void setUltracodeMode: (on: boolean) => void setUpdateStatus: (s: UpdateStatus) => void skillsLoader?: SkillsLoader pluginManager?: PluginManager /** i18n translate function — populated from React tree via useI18n(). * Falls back to module-level defaultT (en-US) in tests / bootstrap. */ t: (key: string, params?: Record) => string } export interface CommandResult { /** System message to display */ content: string /** If true, exit the app */ exit?: boolean /** If provided, update provider after command */ nextProvider?: string /** If provided, update model after command */ nextModel?: string /** If true, clear all messages (handled by caller) */ clearMessages?: boolean /** Content to copy to clipboard (handled by caller) */ copyContent?: string /** If set, route this message to the AI engine — bridges slash commands → skill/agent invocation. * The app layer replaces the user input with this string and falls through to AI processing. */ forwardToAI?: string /** If set, restored messages to load into the session (used by /resume last). */ forwardedMessages?: Message[] /** If true, resume into current non-empty session — app should warn before loading */ resumeWarning?: boolean /** API Key 补录请求：切换前需用户输入 Key */ needsApiKey?: { providerId: string modelId: string providerName: string } } export type CommandHandler = ( ctx: CommandContext, args: string[], ) => CommandResult | Promise // ═══════════════════════════════════════════════════════════════ // Session & Identity // ═══════════════════════════════════════════════════════════════ const helpCmd: CommandHandler = (ctx) => { const skillCount = ctx.skillsLoader?.list().length ?? 14 const toolsCount = ctx.engine.getTools().size const cmdCount = getCommandNames().length return { content: stripIndent` Mipham Code v${PACKAGE_VERSION} — Commands ── Session ────────────────────────── /help Show this help /commands List all ${cmdCount} commands /version Show version info /clear Clear conversation /compact Compact context window /context Show context stats /status Session and system status /cost Token usage estimate /usage Detailed usage dashboard /rename Rename current session /goal [--decompose] [--verify-script ] [--verify-skill ] Set session goal with verification /recap Summarize session so far /summary Generate session summary /stats Session usage statistics /files List files in CWD /cd Change working directory /export Export conversation to file /doctor System diagnostics /resume List saved sessions /resume last Restore most recent session /resume delete Delete a saved session /branch Fork conversation ── History ───────────────────────── /rewind Undo last AI turn /undo Same as /rewind /copy [N] Copy last response to clipboard /focus Toggle focus view (last exchange only) ── Model & Provider ──────────────── /pick Open model picker (or Ctrl+P) /model Show current model /models List all available models /provider Show current provider /providers List configured providers /switch

Switch provider and model /config View configuration /fast [on|off] Toggle fast mode /effort Set reasoning effort (low|medium|high|xhigh|max) /theme [dark|light|auto] Set terminal theme /ollama-refresh Refresh Ollama model list at runtime ── Tools & Skills ────────────────── /tools List available tools (${toolsCount} total) /skills List loaded skills (${skillCount} built-in) /reload-skills Reload all skills /browse-skills Browse community skill marketplace /install-skill Install a skill by name or URL /remove-skill Remove an installed skill /commands List all slash commands /mcp MCP server status /plugins List installed plugins /browse-plugins Browse community plugins /install-plugin Install plugin from npm or path /remove-plugin Remove an installed plugin /plugin-enable Enable a disabled plugin /plugin-disable Disable an enabled plugin ── Workflow ──────────────────────── /plan Enter plan mode (read-only) /no-plan Exit plan mode /tdd [goal] Test-Driven Development workflow /todos [list|create] Task management /tasks Background tasks /review Code review workflow /pr-comments PR review summary /diff Show git diff /workflows List workflow scripts /loop

Run prompt on interval /loop auto

Autonomous self-paced loop /loop stop Stop active autonomous loop /loop init Scaffold .mipham/ vault structure /loop status Show autonomous loop progress /batch Apply changes across files /hooks Manage lifecycle hook scripts /schedule View scheduled tasks ── Git & GitHub ───────────────────── /commit Review staged changes + commit /push Push current branch /pr Create a pull request /issue File a GitHub issue ── Code Quality ───────────────────── /simplify Review for reuse, quality, simplification /lint Run project linter ── Project ───────────────────────── /init Initialize .mipham config /setup Guided project setup wizard /recommend Analyze project + recommend setup /permissions Show or persist permission rules /add-dir

Add workspace directory /security Security review checklist /audit Same as /security ── Environment ───────────────────── /upgrade Show upgrade instructions /release-notes View version changelog /ide IDE integration guide /terminal-setup Shell & terminal config /memory Manage AI memories ── Account ───────────────────────── /login Show API key status /logout Clear credentials guide /keys List API key rotation status /keys rotate Rotate an API key /keys audit Check for expired keys /feedback Send feedback ── Agents ────────────────────────── /agents Agent view dashboard /bg Run a background agent task Type /exit or Esc to quit. `, } } const versionCmd: CommandHandler = (ctx) => { const skillCounts = ctx.skillsLoader?.countByType() const skillsLine = skillCounts ? `Skills: ${skillCounts.total} built-in (${skillCounts.standard} standard + ${skillCounts.mipham} mipham)` : `Skills: (loader unavailable)` const toolsCount = ctx.engine.getTools().size return { content: stripIndent` Mipham Code v${ctx.version} Runtime: Bun ${typeof Bun !== 'undefined' ? Bun.version : '(Node.js)'} Platform: ${process.platform} ${process.arch} Node: ${process.version} CWD: ${process.cwd()} Provider: ${ctx.providerId} / ${ctx.modelId} Tools: ${toolsCount} built-in ${skillsLine} License: Apache 2.0 `, } } const clearCmd: CommandHandler = (ctx) => { ctx.engine.getContext().clear() const t = resolveT(ctx) return { content: t('commands.clear.confirmed'), clearMessages: true } } const exitCmd: CommandHandler = () => ({ content: '', exit: true }) // ═══════════════════════════════════════════════════════════════ // Context & Status // ═══════════════════════════════════════════════════════════════ const compactCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const context = ctx.engine.getContext() const msgsBefore = context.getMessageCount() const tokensBefore = context.getEstimatedTokens() // Show start progress process.stderr.write( `🔄 Compacting: ${msgsBefore} messages, ${tokensBefore.toLocaleString()} tokens...\n`, ) const startTime = Date.now() const result = await context.compact('user requested compaction') const elapsed = ((Date.now() - startTime) / 1000).toFixed(1) const msgsAfter = context.getMessageCount() const saved = result.before > 0 ? ((1 - result.after / result.before) * 100).toFixed(0) : '0' return { content: [ t('commands.compact.confirmed'), `Messages: ${msgsBefore} → ${msgsAfter} | Tokens: ${result.before.toLocaleString()} → ${result.after.toLocaleString()} (${saved}% saved)`, `Duration: ${elapsed}s`, msgsBefore > msgsAfter ? `💡 Tip: Use /compact early to prevent automatic compaction delays.` : '', ] .filter(Boolean) .join('\n'), } } const contextCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const tokens = c.getEstimatedTokens() const msgs = c.getMessages() const systemPromptLen = c.getSystemPrompt().length // 窗口与阈值一律问引擎：模型注册表里 1M/256K/128K/32K 都有，写死 200K 会 // 让 1M 模型显示约 5 倍偏高的百分比，并把压缩点说成 90%（实际 95%）。 const maxTokens = c.getMaxTokens() const threshold = c.getCompactionThreshold() return { content: stripIndent` ${t('commands.context.title')} ${t('commands.context.messages')} ${msgs.length} ${t('commands.context.estimated_tokens')} ${tokens.toLocaleString()} / ${maxTokens.toLocaleString()} ${t('commands.context.usage_pct')} ${((tokens / maxTokens) * 100).toFixed(1)}% ${t('commands.context.system_prompt')} ${systemPromptLen.toLocaleString()} chars (~${Math.ceil(systemPromptLen / 4).toLocaleString()} tokens) ${t('commands.context.compaction')} at ${(threshold * 100).toFixed(0)}% (${Math.round(maxTokens * threshold).toLocaleString()} tokens) `, } } const statusCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const tools = ctx.engine.getTools() const runtime = typeof Bun !== 'undefined' ? 'Bun' : 'Node.js' const runtimeVer = typeof Bun !== 'undefined' ? Bun.version : process.version const health = ctx.engine.getRegistry() ? await ctx.engine.getRegistry().healthMap() : new Map() const providerHealth = ctx.config.providers .filter((p) => p.status !== 'upcoming') .map((p) => `${p.id} ${health.get(p.id) ? '🟢' : '🔴'}`) .join(' ') return { content: stripIndent` ${t('commands.status.session_title')} ${t('commands.status.provider')} ${ctx.providerId} ${t('commands.status.model')} ${ctx.modelId} ${t('commands.status.messages')} ${c.getMessages().length} ${t('commands.status.tokens')} ~${c.getEstimatedTokens().toLocaleString()} / ${c.getMaxTokens().toLocaleString()} ${t('commands.status.tools')} ${tools.size} ${t('commands.status.loaded')} ${t('commands.status.permission')} ${ctx.config.permission} ${t('commands.status.system_title')} ${t('commands.status.platform')} ${process.platform} ${process.arch} ${t('commands.status.runtime')} ${runtime} ${runtimeVer} ${t('commands.status.cwd')} ${process.cwd()} Providers ${providerHealth} `, } } const costCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const tokens = c.getEstimatedTokens() const cacheStatus = c.getCacheStatus() const cachedTokens = cacheStatus.cachedTokens const hitRatio = tokens > 0 ? Math.min(1, cachedTokens / tokens) : 0 const uncachedTokens = Math.max(0, tokens - cachedTokens) const maxTokens = c.getMaxTokens() return { content: stripIndent` ${t('commands.context_tokens.title')} ${t('commands.context_tokens.context_tokens')} ~${tokens.toLocaleString()} / ${maxTokens.toLocaleString()} ${t('commands.context_tokens.usage')} ${((tokens / maxTokens) * 100).toFixed(1)}% ${t('commands.context_tokens.prompt_cache', { cached: cachedTokens.toLocaleString(), ratio: (hitRatio * 100).toFixed(1), uncached: uncachedTokens.toLocaleString(), })} ${t('commands.context_tokens.footer')} `, } } // ═══════════════════════════════════════════════════════════════ // Model & Provider // ═══════════════════════════════════════════════════════════════ const modelCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) return { content: t('commands.model.current', { model: ctx.modelId, provider: ctx.providerId }) } } const modelsCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const lines = ctx.config.providers .filter((p) => p.status !== 'upcoming') .flatMap((p) => p.models .filter((m) => m.status === 'active') .map( (m) => ` ${p.id.padEnd(12)} ${m.id.padEnd(30)} ${formatContextWindow(m.contextWindow)} ctx ${m.vision ? '🖼' : '📝'}`, ), ) return { content: `${t('commands.models.title', { count: String(lines.length) })}\n\n${t('commands.models.header')}\n${'-'.repeat(80)}\n${lines.join('\n')}\n\n${t('commands.models.hint')}`, } } const ollamaRefreshCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const ollamaProvider = ctx.config.providers.find((p) => p.id === 'ollama') if (!ollamaProvider) { return { content: t('commands.ollama_refresh.not_configured') } } const before = ollamaProvider.models.length try { const seen = new Set() const refreshed: Array<{ id: string }> = [] // 1. Scan locally downloaded models via `ollama list` try { const out = execSync('ollama list', { timeout: 5000, encoding: 'utf-8' }) const lines = out.split('\n').slice(1).filter(Boolean) for (const line of lines) { const name = line.split(/\s+/)[0]! if (!seen.has(name)) { seen.add(name) refreshed.push({ id: name }) } } } catch { // ollama list failed (not installed / not running) — continue with presets } // 2. Merge preset models (deduplicated) for (const preset of OLLAMA_PRESET_MODELS) { if (!seen.has(preset.id)) { seen.add(preset.id) refreshed.push({ id: preset.id }) } } // 3. Update the in-memory provider model list ollamaProvider.models = refreshed.map((m) => ({ id: m.id, name: m.id, providerId: 'ollama', contextWindow: 128_000, maxOutput: 32_000, vision: false, status: 'active' as const, })) const after = refreshed.length const added = after - before return { content: t('commands.ollama_refresh.done', { total: String(after), added: added > 0 ? `+${added}` : String(added), }), } } catch (err: unknown) { const msg = err instanceof Error ? err.message : String(err) return { content: t('commands.ollama_refresh.error', { error: msg }) } } } const providerCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) return { content: t('commands.provider.current', { provider: ctx.providerId, model: ctx.modelId }), } } const providersCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const health = ctx.engine.getRegistry() ? await ctx.engine.getRegistry().healthMap() : new Map() const lines = ctx.config.providers.map((p) => { if (p.status === 'upcoming') { return ` ${p.id.padEnd(14)} ${p.name.padEnd(20)} ${p.protocol.padEnd(18)} ${p.models.length} models ⏳ [upcoming]` } const ok = health.get(p.id) ?? false return ` ${p.id.padEnd(14)} ${p.name.padEnd(20)} ${p.protocol.padEnd(18)} ${p.models.length} models ${ok ? '🟢 reachable' : '🔴 unreachable'}` }) return { content: `${t('commands.providers.title')}\n\n${lines.join('\n')}\n\n${t('commands.providers.current', { provider: ctx.providerId, model: ctx.modelId })}`, } } /** Detect whether an API key is missing (empty, placeholder, or env-var reference that resolves to empty). */ function isApiKeyMissing(apiKey: string): boolean { if (!apiKey || apiKey.trim() === '') return true if (apiKey === 'ollama-local') return false // Ollama local — no key needed if (/^\$\{[A-Z_]+\}$/.test(apiKey.trim())) return true // ${VAR} placeholder return false } const switchCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const [newProvider, newModel] = args if (!newProvider || !newModel) { return { content: t('commands.switch.usage') } } const provider = ctx.config.providers.find((p) => p.id === newProvider) if (!provider) { return { content: t('commands.switch.unknown_provider', { provider: newProvider }) } } // Check API Key before switching if (isApiKeyMissing(provider.apiKey)) { return { content: t('commands.switch.needs_api_key', { provider: provider.name }), needsApiKey: { providerId: newProvider, modelId: newModel, providerName: provider.name }, } } ctx.engine.switchProvider(newProvider, newModel) return { content: t('commands.switch.confirmed', { provider: newProvider, model: newModel }), nextProvider: newProvider, nextModel: newModel, } } const configCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.config const xs = loadCrossSessionConfig(process.cwd()) const lines = [ `version: ${c.version}`, `defaultProvider: ${c.defaultProvider}`, `defaultModel: ${c.defaultModel}`, `permission: ${c.permission}`, `showThinking: ${c.showThinking ?? 'off'}`, `showSchedulingNotices: ${c.showSchedulingNotices ?? false}`, `showCommandPicker: ${c.showCommandPicker ?? false}`, `providers: ${c.providers.length} configured`, '', `crossSessionInbound: ${xs.crossSessionInbound}`, `dialogExpiry: ${xs.dialogExpiry}s`, ] return { content: `${t('commands.config.title')}\n${lines.join('\n')}\n\n${t('commands.config.edit')}`, } } // ═══════════════════════════════════════════════════════════════ // Tools & Skills // ═══════════════════════════════════════════════════════════════ const toolsCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const tools = ctx.engine.getTools() const categories: Record = { file: [], exec: [], agent: [], network: [], system: [], } for (const [name, tool] of tools) { categories[tool.category]?.push( ` ${name.padEnd(14)} ${tool.permission.padEnd(8)} ${tool.description}`, ) } const sections = Object.entries(categories) .filter(([, v]) => v!.length > 0) .map(([cat, items]) => `── ${cat.toUpperCase()} ──\n${items!.join('\n')}`) return { content: `${t('commands.tools.title', { count: String(tools.size) })}\n\n${sections.join('\n\n')}`, } } const skillsCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) if (!ctx.skillsLoader) { return { content: t('commands.skills.unavailable') } } const counts = ctx.skillsLoader.countByType() const standard = ctx.skillsLoader.listByType('standard') const mipham = ctx.skillsLoader.listByType('mipham') const lines: string[] = [ `── Standard Skills (${counts.standard}) ──`, ...standard.map((s) => ` ${s.name.padEnd(28)} ${s.description}`), '', `── Mipham Exclusive (${counts.mipham}) ──`, ...mipham.map((s) => ` ${s.name.padEnd(20)} ${s.description}`), '', t('commands.skills.loaded', { count: String(counts.total) }), ] return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // CRSI (Conversational Rule Self-Improvement) // ═══════════════════════════════════════════════════════════════ const crsiRulesCmd: CommandHandler = (ctx) => { const engine = ctx.engine.getRuleEngine() if (!engine) { return { content: 'CRSI rule engine is not available.' } } const rules = engine.getActiveRules() if (rules.length === 0) { return { content: 'No active CRSI rules.' } } const lines: string[] = ['## Active CRSI Rules', ''] for (const r of rules) { const status = r.enabled ? '✅' : '⛔' lines.push(`- \`${r.id}\` [${r.category}] ${r.toolName} — ${r.source} ${status}`) } lines.push('', `Total: ${rules.length} active rules`) lines.push('', 'Use `/crsi disable ` to disable a rule.') return { content: lines.join('\n') } } const crsiDisableCmd: CommandHandler = (ctx, args) => { const engine = ctx.engine.getRuleEngine() if (!engine) { return { content: 'CRSI rule engine is not available.' } } const ruleId = args[0]?.trim() if (!ruleId) { return { content: 'Usage: /crsi disable ' } } engine.setRuleEnabled(ruleId, false) return { content: `Rule \`${ruleId}\` has been disabled. Use \`/crsi restore ${ruleId}\` to re-enable.`, } } const crsiAnalyzeCmd: CommandHandler = async (ctx) => { const analyzer = ctx.engine.getPatternAnalyzer() const engine = ctx.engine.getRuleEngine() if (!engine) { return { content: 'CRSI system is not available.' } } const patterns = analyzer.analyzeAllAgents() if (patterns.length === 0) { return { content: 'No failure patterns found across agents.' } } let registered = 0 for (const pattern of patterns) { const toolRule = analyzer.toToolRule(pattern) engine.register(toolRule) registered++ } const lines: string[] = [ '## CRSI Analysis Complete', '', `Found ${patterns.length} patterns, ${registered} rules registered.`, '', ] for (const p of patterns) { lines.push( `- [${p.category}] \`${p.agentName}\` — ${p.frequency} failures (${p.confidence} confidence)`, ) } return { content: lines.join('\n') } } const crsiRestoreCmd: CommandHandler = (ctx, args) => { const engine = ctx.engine.getRuleEngine() if (!engine) { return { content: 'CRSI rule engine is not available.' } } const ruleId = args.join(' ').trim() if (!ruleId) { return { content: 'Usage: /crsi restore ' } } engine.setRuleEnabled(ruleId, true) return { content: `Rule \`${ruleId}\` has been re-enabled.` } } /** Render a tiny ASCII sparkline from a series of 0-1 values (success rates). */ function crsiSparkline(values: number[]): string { if (values.length === 0) return '—' const chars = '▁▂▃▄▅▆▇█' return values.map((v) => chars[Math.min(Math.max(Math.round(v * 7), 0), 7)] ?? '▁').join('') } const crsiStatsCmd: CommandHandler = async (ctx) => { const engine = ctx.engine.getRuleEngine() const tracker = ctx.engine.getEffectivenessTracker() if (!engine) { return { content: 'CRSI rule engine is not available.' } } const rules = engine.getActiveRules() const lines: string[] = ['## CRSI Statistics', ''] lines.push(`Total active rules: ${rules.length}`) lines.push(`Builtin: ${rules.filter((r) => r.source === 'builtin').length}`) lines.push(`Auto-generated: ${rules.filter((r) => r.source === 'pattern-analyzer').length}`) lines.push(`Manual: ${rules.filter((r) => r.source === 'manual').length}`) const effs = tracker?.allRules ?? [] if (effs.length > 0) { let totalInterceptions = 0 let totalSuccesses = 0 for (const eff of effs) { totalInterceptions += eff.appliedCount totalSuccesses += eff.successAfterCount } lines.push('') lines.push(`Total interceptions: ${totalInterceptions}`) lines.push( `Success rate after rules: ${totalInterceptions > 0 ? Math.round((totalSuccesses / totalInterceptions) * 100) : 0}%`, ) lines.push('') lines.push('### Rule effectiveness') lines.push('| Rule | Status | Applied | Success | Trend |') lines.push('|------|--------|---------|---------|-------|') const statusIcon: Record = { active: '🟢', degrading: '🟡', disabled: '⚫' } for (const eff of effs) { const icon = statusIcon[eff.status] ?? '⚪' const rate = eff.appliedCount > 0 ? Math.round((eff.successAfterCount / eff.appliedCount) * 100) : 0 const trend = crsiSparkline(eff.evaluationHistory.map((h) => 1 - h.failureRate)) lines.push( `| ${eff.ruleId} | ${icon} ${eff.status} | ${eff.appliedCount} | ${rate}% | ${trend} |`, ) } const disabled = effs.filter((e) => e.status === 'disabled') if (disabled.length > 0) { lines.push('') lines.push('### Disabled rules (retired)') for (const d of disabled) { lines.push(`- ${d.ruleId} (failure rate ${Math.round(d.postRuleFailureRate * 100)}%)`) } } } // ── ε 预测命中（prose 路径） ── // 作废条款：样本不足就不下结论；台账攒够 20 条而判定样本仍 < 5 ⇒ 明写机制失效。 // 只打印、不写台账：本命令是只读的，而该结论每次都能从同一份 improvements.jsonl 重算出来。 const records = readImprovements() const pred = predictionHitRate(records) lines.push('') lines.push('### ε 预测命中（prose 路径）') if (pred.total < 5) { lines.push(`样本不足（判定记录 ${pred.total} 条，需 ≥ 5）—— 不下结论。`) if (records.length >= 20) { lines.push('⚠️ ε 机制失效：prose 路径使用率过低（记录总数已达 20 而判定样本仍 < 5）。') } } else { lines.push( `命中率: ${pred.hits}/${pred.total} (${(pred.rate * 100).toFixed(0)}%, ` + `Wilson 95% [${(pred.lo * 100).toFixed(0)}%, ${(pred.hi * 100).toFixed(0)}%])`, ) } return { content: lines.join('\n') } } const crsiInventoryCmd: CommandHandler = async (ctx) => { return { content: buildCapabilityReport(ctx.engine) } } const crsiModifyCmd: CommandHandler = async (ctx, args) => { if (args[0] === '--approve') { if (shouldBlockApproval(getPendingVerdict() ?? 'inconclusive')) { return { content: '❌ 任务表现倒退，禁止固化。请 /crsi modify --reject 丢弃，或改进后再试。' } } const r = approvePending() setPendingVerdict(null) return { content: r.success ? `✅ ${r.message}` : `⚠️ ${r.message}` } } if (args[0] === '--reject') { const r = rejectPending() setPendingVerdict(null) return { content: r.success ? `✅ ${r.message}` : `⚠️ ${r.message}` } } if (args.length < 3) { return { content: 'Usage: /crsi modify \n' + '- description 单 token（不含空格）\n' + '- newContent 用 \\n 表示换行\n' + '测试通过后：/crsi modify --approve 合并，/crsi modify --reject 丢弃', } } if (hasPending()) { return { content: '⚠️ 已有待批准的修改。先 /crsi modify --approve 或 --reject。' } } const description = args[0]! const filePath = args[1]! const newContent = args.slice(2).join(' ').replace(/\\n/g, '\n') let originalContent = '' try { const { readFileSync } = await import('node:fs') const { join } = await import('node:path') originalContent = readFileSync(join(process.cwd(), filePath), 'utf-8') } catch { // 文件不存在 → 宽松模式（originalContent 为空） } const result = await runCrsiModification({ description, filePath, newContent, originalContent, blastRadius: [filePath], }) if (!result.applied || result.phase === 'failed') { return { content: `❌ 修改未通过（phase: ${result.phase}）。\n${result.error ?? ''}`, } } // 测量在 runCrsiModification 成功后进行（避免 failed proposal 白跑 6 次 LLM 调用）。 const llm = ctx.engine.getLlm() ?? ctx.engine.getRegistry() let improvementLine = '' try { const sample = await measureSkillDeltaRepeated(llm, { filePath, originalContent, newContent }) if (sample) { const report = buildImprovementReport(sample, [filePath]) setPendingVerdict(report.verdict) appendImprovement({ ...report, id: randomUUID(), timestamp: new Date().toISOString() }) const rate = improvementRate(readImprovements()) const label = report.verdict === 'improved' ? 'improved ✅' : report.verdict === 'regressed' ? 'regressed ⚠️' : 'inconclusive' const sign = report.deltaMean >= 0 ? '+' : '' // B2 代价维：只展示、不进判定（`formatCostLine` 缺席时返回 null ⇒ 整行不打）。 const costLine = formatCostLine(report) improvementLine = `\n📊 改进判定: ${label} (delta ${sign}${report.deltaMean.toFixed(1)}, 噪声 ${report.noise.toFixed(1)}, 阈值 ${report.minEffect.toFixed(1)})` + `\n 改进率: ${rate.improved}/${rate.total} (${(rate.rate * 100).toFixed(0)}%, Wilson 95% [${(rate.lo * 100).toFixed(0)}%, ${(rate.hi * 100).toFixed(0)}%])` + (costLine ? `\n${costLine}` : '') + (report.verdict === 'regressed' ? '\n ⚠️ 任务表现倒退：--approve 将被拒绝。' : '') } } catch { // 测量失败（LLM 不可用等）不阻断 modify 流程——改进信号是可选的。 } return { content: `✅ 测试通过。审阅下方 diff：\n\n${result.diff}\n` + improvementLine + '\n/crsi modify --approve 合并\n/crsi modify --reject 丢弃', } } const crsiProposeCmd: CommandHandler = async (ctx, args) => { if (hasPending()) { return { content: '⚠️ 已有待批准的修改。先 /crsi modify --approve 或 --reject。' } } const insights = ctx.engine.getAutoMemory?.()?.accumulatedInsights ?? [] let metaRules: Parameters[1] = [] try { metaRules = ctx.engine.getMetaRuleEngine?.()?.analyze().metaRules ?? [] } catch { metaRules = [] } // 目标文件按仓库根解析（沙箱的 filePath 是仓库根相对）。 let root = process.cwd() try { root = execSync('git rev-parse --show-toplevel', { timeout: 5000, encoding: 'utf-8', }).trim() } catch { // 非 git 目录 → 回退 cwd } // ── 散文提议路径：/crsi propose --prose 用 LLM 生成改 skill 散文提议 ── if (args[0] === '--prose') { const skillFiles = collectSkillFiles(root) if (skillFiles.length === 0) { return { content: '没有可改的 skill 文件。' } } const signal = selectCrsiSignal(insights, metaRules) if (!signal) { return { content: '没有足够的失败信号来生成散文提议。' } } const id = proseProposalId(signal) if (hasProposedProse(id)) { return { content: '该失败信号已生成过散文提议（幂等去重，跳过）。' } } const llm = ctx.engine.getLlm() ?? ctx.engine.getRegistry() const { readFileSync } = await import('node:fs') const { join } = await import('node:path') const proposal = await produceProseProposal(signal, llm, skillFiles, (p) => readFileSync(join(root, p), 'utf-8'), ) if (!proposal) { return { content: '散文提议生成失败（LLM 未返回有效结果）。' } } const verdict = prefilterProposal({ id, filePath: proposal.filePath, kind: 'skill', newContent: proposal.newContent, }) if (!verdict.pass) { return { content: `❌ 提议未通过预筛：${verdict.reasons.join('; ')}` } } const result = await runCrsiModification({ description: proposal.description, filePath: proposal.filePath, newContent: proposal.newContent, originalContent: proposal.originalContent, blastRadius: [proposal.filePath], }) if (!result.applied || result.phase === 'failed') { return { content: `❌ 生成失败（phase: ${result.phase}）。\n${result.error ?? ''}` } } // ε 预登记落地：prose 路径此前**不测量**（measureSkillDeltaRepeated 全仓库只有手工路径一个 // 调用点）⇒ ε 曾在 A 流程登记、判定侧在 B 流程，两端永不相遇。这里补上测量。 // 成本：每次提案多 6 次 LLM 调用（同手工路径 :867 的注释）。 let predictionLine = '' try { const sample = await measureSkillDeltaRepeated(llm, { filePath: proposal.filePath, originalContent: proposal.originalContent, newContent: proposal.newContent, }) if (sample) { const report = buildImprovementReport(sample, [proposal.filePath], proposal.expectedEffect) setPendingVerdict(report.verdict) appendImprovement({ ...report, id: randomUUID(), timestamp: new Date().toISOString() }) if (report.predictionHit !== undefined) { predictionLine = `\n🎯 ε 预测命中: ${report.predictionHit ? '命中 ✅' : '未命中 ⚠️'}` + `（预测 ${report.predictedDelta}，实际 delta ${report.deltaMean.toFixed(1)}）` } // B2 代价维：**与手工路径同一行读数**，两条渲染路径都接（只接一条即本仓库记过的 // 「局部正确全局遗漏」）。ε 行可以有、代价行可以没有，故各自独立追加。 const costLine = formatCostLine(report) if (costLine) predictionLine += `\n${costLine}` } } catch { // 测量失败（LLM 不可用等）不阻断提案流程 —— 与手工路径 :889 的处置一致。 } appendProseProposal({ id, filePath: proposal.filePath, timestamp: new Date().toISOString() }) return { content: `✅ 已生成散文提议并跑过测试。审阅 diff：\n\n${result.diff}\n\n` + predictionLine + '/crsi modify --approve 合并 | /crsi modify --reject 丢弃', } } // ── 毕业路径：/crsi propose --rule 固化受管理规则（行为） ── if (args[0] === '--rule') { const signal = selectCrsiSignal(insights, metaRules) if (!signal) { return { content: '没有足够的失败信号（autoApplicable insight 或高置信元规则）来固化规则。' } } let current = '' try { const { readFileSync } = await import('node:fs') const { join } = await import('node:path') current = readFileSync(join(root, MANAGED_RULES_FILE), 'utf-8') } catch { current = '' } const proposal = produceRuleProposal(signal, current) if (!proposal) { return { content: '没有可固化的规则（category 需为 timeout/tool-params，且同名规则不存在）。', } } const result = await runCrsiModification(proposal) if (!result.applied || result.phase === 'failed') { return { content: `❌ 固化失败（phase: ${result.phase}）。\n${result.error ?? ''}` } } return { content: `✅ 已生成受管理规则并跑过测试。审阅 diff：\n\n${result.diff}\n\n` + '/crsi modify --approve 合并 | /crsi modify --reject 丢弃', } } // ── Crossover 路径：/crsi propose --crossover 合并两条重叠教训 ── if (args[0] === '--crossover') { const llm = ctx.engine.getLlm() ?? ctx.engine.getRegistry() let current = '' try { const { readFileSync } = await import('node:fs') const { join } = await import('node:path') current = readFileSync(join(root, LESSONS_FILE), 'utf-8') } catch { current = '' } if (!current) { return { content: '教训文件为空，无可合并。' } } const proposal = await produceCrossoverProposal(llm, current, new Date().toISOString()) if (!proposal) { return { content: '没有找到可合并的重叠教训对。' } } const result = await runCrsiModification(proposal) if (!result.applied || result.phase === 'failed') { return { content: `❌ 合并失败（phase: ${result.phase}）。\n${result.error ?? ''}` } } return { content: `✅ 已生成合并教训并跑过测试。审阅 diff：\n\n${result.diff}\n\n` + '/crsi modify --approve 合并 | /crsi modify --reject 丢弃', } } // ── 教训路径（默认）：/crsi propose 追加教训 ── let current = '' try { const { readFileSync } = await import('node:fs') const { join } = await import('node:path') current = readFileSync(join(root, LESSONS_FILE), 'utf-8') } catch { current = '' } const proposal = produceCrsiProposal(insights, metaRules, current, new Date().toISOString()) if (!proposal) { return { content: '没有足够的失败信号（autoApplicable insight 或高置信元规则）来生成教训。' } } const result = await runCrsiModification(proposal) if (!result.applied || result.phase === 'failed') { return { content: `❌ 生成失败（phase: ${result.phase}）。\n${result.error ?? ''}` } } return { content: `✅ 已生成教训并跑过测试。审阅 diff：\n\n${result.diff}\n\n` + '/crsi modify --approve 合并 | /crsi modify --reject 丢弃', } } const crsiEvalCmd: CommandHandler = async (ctx, args) => { const rewardIdx = args.indexOf('--reward') const rewardName = rewardIdx >= 0 ? args[rewardIdx + 1] : undefined if (rewardName) { const llm = ctx.engine.getLlm() ?? ctx.engine.getRegistry() const fns = listRewardFns(llm) const fn = fns.find((f) => f.name === rewardName) if (!fn) { return { content: `❌ 未知 reward: ${rewardName}。可用: ${fns.map((f) => f.name).join(', ')}`, } } // 先读后写：比较基准是「上一次已落盘的态」，不依赖刚写进去那条排第几。 const prev = getContractHistory(fn.name) const report = await fn.evaluate() const deltaLines = report.results ? renderContractDiff(diffContractHistory(report.results, prev)) : [] appendEvalScore(fn.name, report) return { content: [ `得分 **${report.score}/100** (${report.passed}/${report.total})`, `失败: ${report.failures.join(', ') || '无'}`, ...(deltaLines.length > 0 ? ['', '### 与上次相比', ...deltaLines] : []), ].join('\n'), } } // 先读后写（同上）：基准是上一次落盘的态。 const prevSnapshot = getContractHistory('mechanism-sentinel') const report = runEval() const deltaLines = renderContractDiff(diffContractHistory(report.results, prevSnapshot)) appendEvalScore('mechanism-sentinel', report) const lines: string[] = ['## 🧪 CRSI Eval Harness', ''] lines.push(`得分: **${report.score}/100** (${report.passed}/${report.total})`, '') lines.push('| 任务 | 结果 |') lines.push('|------|------|') for (const r of report.results) { lines.push( `| ${r.description} | ${r.passed ? '✅' : '❌'}${r.detail ? ` — ${r.detail}` : ''} |`, ) } if (report.failures.length > 0) { lines.push('', `❌ 失败任务: ${report.failures.join(', ')}`) } // 只读展示：账本按契约粒度落盘后，才能回答「是哪条契约翻的」。 // 这里不做任何自动决策——闸门仍只看 regressedAnchors / score。 if (deltaLines.length > 0) { lines.push('', '### 与上次相比', ...deltaLines) } // 奖励函数注册表（reward function = policy→feedback 抽象可见） // 传 llm 列出完整注册表（task-performance 需 llm 才能跑，但构造它零 LLM 调用）。 const llm = ctx.engine.getLlm() ?? ctx.engine.getRegistry() const fns = listRewardFns(llm) lines.push('', '## 🎁 奖励函数注册表', '') for (const f of fns) { lines.push(`- **${f.name}** — ${f.description}`) } lines.push('', '`/crsi eval --reward ` 跑指定奖励函数') return { content: lines.join('\n') } } const crsiBenchCmd: CommandHandler = async (ctx, args) => { const llm = ctx.engine.getLlm() ?? ctx.engine.getRegistry() const skillIdx = args.indexOf('--skill') const skillName = skillIdx >= 0 ? args[skillIdx + 1] : undefined let skill: { name: string; text: string } | undefined if (skillName) { const body = ctx.skillsLoader?.get(skillName)?.body if (!body) { return { content: `❌ 未找到 skill: ${skillName}` } } skill = { name: skillName, text: body } } const report = await runTaskPerformance(llm, skill ? { skill } : undefined) const lines: string[] = [ '## 🎯 CRSI 任务表现基准' + (skill ? `（skill: ${skill.name}）` : ''), '', ] lines.push(`得分: **${report.score}/100** (${report.passed}/${report.total})`, '') lines.push('| 任务 | 结果 |') lines.push('|------|------|') for (const r of report.results) { lines.push( `| ${r.description.slice(0, 60)} | ${r.passed ? '✅' : '❌'}${r.detail ? ` — ${r.detail.slice(0, 80)}` : ''} |`, ) } if (report.failures.length > 0) { lines.push('', `❌ 失败任务: ${report.failures.join(', ')}`) } return { content: lines.join('\n') } } const crsiProseClearCmd: CommandHandler = () => { const count = clearProseProposals() if (count === 0) { return { content: '散文提议 ledger 为空（无可清除记录）。' } } return { content: `已清空散文提议 ledger（移除 ${count} 条记录）。` } } const crsiHealthCmd: CommandHandler = async (ctx) => { const engine = ctx.engine.getRuleEngine() const tracker = ctx.engine.getEffectivenessTracker() const db = ctx.engine.getErrorSignatureDB?.() const patternAnalyzer = ctx.engine.getPatternAnalyzer?.() const lines: string[] = ['## 🏥 CRSI 系统健康仪表盘', ''] // ── CRSI Section ── lines.push('### 🧠 CRSI 学习子系统', '') if (engine) { const rules = engine.getActiveRules() const builtin = rules.filter((r) => r.source === 'builtin').length const auto = rules.filter((r) => r.source === 'pattern-analyzer').length const manual = rules.filter((r) => r.source === 'manual').length lines.push(`| 指标 | 值 |`) lines.push(`|------|----|`) lines.push(`| 活跃规则 | ${rules.length} (内置 ${builtin} · 自动 ${auto} · 手动 ${manual}) |`) if (tracker) { let interceptions = 0 let successes = 0 for (const r of rules) { const eff = tracker.getEffectiveness(r.id) if (eff) { interceptions += eff.appliedCount successes += eff.successAfterCount } } const crsiRate = interceptions > 0 ? Math.round((successes / interceptions) * 100) : 100 lines.push(`| 总拦截次数 | ${interceptions} |`) lines.push(`| 拦截成功率 | ${crsiRate}% |`) } // Pattern analysis depth if (patternAnalyzer) { const patterns = patternAnalyzer.analyzeAllAgents() lines.push(`| 已检测模式 | ${patterns.length} |`) } lines.push('') lines.push('CRSI 命令: `/crsi rules` `/crsi stats` `/crsi analyze`') } else { lines.push('CRSI 规则引擎未初始化') } // ── SIS Section ── lines.push('', '### 🛡️ SIS 自免疫子系统', '') if (db) { const stats = db.getStats() lines.push(`| 指标 | 值 |`) lines.push(`|------|----|`) lines.push( `| 免疫记忆 | ${stats.total} 条 (🟢${stats.active} · 🟡${stats.degraded} · ⚫${stats.retired}) |`, ) lines.push(`| 平均成功率 | ${Math.round(stats.avgSuccessRate * 100)}% |`) lines.push(`| 总拦截 | ${stats.totalInterceptions} 次 |`) lines.push('') lines.push('SIS 命令: `/sis errors` `/sis stats` `/sis clear `') } else { lines.push('SIS 自免疫系统未初始化') } // ── Overall Health Score ── lines.push('', '### 📊 综合健康评分', '') const crsiScore = engine ? 50 : 0 const sisScore = db ? (db.getStats().avgSuccessRate >= 0.8 ? 50 : 25) : 0 const totalScore = crsiScore + sisScore const grade = totalScore >= 80 ? '🟢 A — 优秀' : totalScore >= 50 ? '🟡 B — 良好' : totalScore >= 25 ? '🟠 C — 需关注' : '🔴 D — 待初始化' lines.push(`| 子系统 | 状态 | 得分 |`) lines.push(`|--------|------|------|`) lines.push(`| CRSI 学习 | ${engine ? '✅ 在线' : '❌ 离线'} | ${crsiScore}/50 |`) lines.push(`| SIS 免疫 | ${db ? '✅ 在线' : '❌ 离线'} | ${sisScore}/50 |`) lines.push(`| **综合** | **${grade}** | **${totalScore}/100** |`) return { content: lines.join('\n') } } const crsiMetaCmd: CommandHandler = async (ctx) => { const metaEngine = ctx.engine.getMetaRuleEngine?.() if (!metaEngine) { return { content: 'MetaRuleEngine 未初始化。请先运行 `/crsi health` 初始化 CRSI 系统。' } } const result = metaEngine.analyze() const lines: string[] = ['## 🔮 CRSI 元规则分析 (RSI Level 3)', ''] // ── System Health ── lines.push('### 📊 系统健康评分', '') const h = result.systemHealth lines.push(`| 组件 | 得分 |`) lines.push(`|------|------|`) lines.push(`| 错误签名 | ${h.components.errorSignatures}/100 |`) lines.push(`| 预防拦截 | ${h.components.preflightPrevention}/100 |`) lines.push(`| 自动修复 | ${h.components.autoCorrection}/100 |`) lines.push(`| 免疫记忆 | ${h.components.immuneMemory}/100 |`) lines.push(`| 规则有效性 | ${h.components.ruleEffectiveness}/100 |`) lines.push(`| **综合** | **${h.score}/100** |`) lines.push('') lines.push(h.assessment) lines.push('') // ── Recommendations ── lines.push('### 💡 优化建议', '') for (const rec of h.recommendations) { lines.push(`- ${rec}`) } lines.push('') // ── Meta-Rules ── if (result.metaRules.length > 0) { lines.push(`### 🧬 发现 ${result.metaRules.length} 条元规则`, '') for (const mr of result.metaRules) { const confIcon = mr.confidence === 'high' ? '🔴' : mr.confidence === 'medium' ? '🟡' : '🟢' const autoLabel = mr.autoApplicable ? ' [可自动应用]' : '' lines.push(`#### ${confIcon} ${mr.title}${autoLabel}`) lines.push('') lines.push(mr.description) lines.push('') lines.push(`> **建议**: ${mr.recommendation}`) lines.push('') lines.push( `置信度: ${mr.confidence} | 样本量: ${mr.evidence.sampleSize} | 生成时间: ${mr.generatedAt}`, ) lines.push('') } } else { lines.push('### 🧬 元规则', '') lines.push( '暂无元规则生成。当系统积累足够的错误签名和规则数据后，元规则引擎将自动发现跨规则模式。', ) lines.push('') } // ── Auto-applicable summary ── const autoApplicable = metaEngine.getAutoApplicable(result.metaRules) if (autoApplicable.length > 0) { lines.push('### ⚡ 可自动应用', '') lines.push(`${autoApplicable.length} 条元规则可安全自动应用：`) for (const mr of autoApplicable) { lines.push(`- \`${mr.id}\`: ${mr.recommendation}`) } lines.push('') lines.push('使用 `/crsi meta --apply` 自动应用高置信度元规则。') } return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // SIS (Self-Immune System) // ═══════════════════════════════════════════════════════════════ const sisErrorsCmd: CommandHandler = (ctx) => { const db = ctx.engine.getErrorSignatureDB?.() if (!db) { return { content: 'SIS 自免疫系统未初始化。' } } const sigs = db.getActive() if (sigs.length === 0) { return { content: '🛡️ SIS 免疫记忆为空 — 尚未记录任何错误签名。' } } const lines: string[] = ['## 🛡️ SIS 免疫记忆', ''] for (const sig of sigs) { const statusIcon = sig.status === 'active' ? '🟢' : '🟡' lines.push(`- ${statusIcon} \`${sig.id}\` [${sig.category}] ${sig.toolName}`) lines.push(` 模式: \`${sig.pattern.slice(0, 80)}${sig.pattern.length > 80 ? '...' : ''}\``) lines.push( ` 修复: ${sig.fixStrategy} → ${sig.fixAction.slice(0, 60)} | 发生 ${sig.occurrences} 次 | 成功率 ${Math.round(sig.successRate * 100)}%`, ) lines.push('') } lines.push(`共 ${sigs.length} 条活跃免疫记忆`) lines.push('', '使用 `/sis stats` 查看汇总统计，`/sis clear ` 清除指定签名') return { content: lines.join('\n') } } const sisStatsCmd: CommandHandler = (ctx) => { const db = ctx.engine.getErrorSignatureDB?.() if (!db) { return { content: 'SIS 自免疫系统未初始化。' } } const stats = db.getStats() const lines: string[] = ['## 🛡️ SIS 自免疫统计', ''] lines.push(`总签名数: ${stats.total}`) lines.push( `🟢 活跃: ${stats.active} | 🟡 降级: ${stats.degraded} | ⚫ 已退役: ${stats.retired}`, ) lines.push(`平均成功率: ${Math.round(stats.avgSuccessRate * 100)}%`) lines.push(`总拦截次数: ${stats.totalInterceptions}`) const active = db.getActive() if (active.length > 0) { lines.push('') lines.push('### 签名明细（按拦截次数排序）') lines.push('| 签名 | 状态 | 拦截 | 成功率 |') lines.push('|------|------|------|--------|') for (const s of active) { const icon = s.status === 'active' ? '🟢' : '🟡' lines.push( `| ${s.id} | ${icon} ${s.status} | ${s.occurrences} | ${Math.round(s.successRate * 100)}% |`, ) } } return { content: lines.join('\n') } } const sisClearCmd: CommandHandler = (ctx, args) => { const db = ctx.engine.getErrorSignatureDB?.() if (!db) { return { content: 'SIS 自免疫系统未初始化。' } } const sigId = args[0]?.trim() if (!sigId) { return { content: 'Usage: /sis clear \n\n使用 `/sis errors` 查看所有签名 ID。' } } const sig = db.get(sigId) if (!sig) { return { content: `签名 \`${sigId}\` 未找到。` } } db.retire(sigId) return { content: `已退役签名 \`${sigId}\` (${sig.pattern.slice(0, 50)}...)` } } const sisCleanupCmd: CommandHandler = async (ctx) => { const db = ctx.engine.getErrorSignatureDB?.() if (!db) { return { content: 'SIS 自免疫系统未初始化。' } } // Use dynamic import for ESM compatibility const { ImmuneMemoryGC } = await import('../../src/core/immune-memory-gc.js') const gc = new ImmuneMemoryGC(db) const report = gc.collect() const lines: string[] = ['## 🧹 SIS 免疫记忆清理完成', ''] lines.push(`清理前: ${report.before} 条签名`) lines.push(`清理后: ${report.after} 条签名`) lines.push('') if (report.retiredRemoved > 0) lines.push(`🗑️ 移除过期退役: ${report.retiredRemoved} 条`) if (report.zeroSuccessRetired > 0) lines.push(`⚠️ 退役零成功率: ${report.zeroSuccessRetired} 条`) if (report.duplicatesMerged > 0) lines.push(`🔗 合并重复签名: ${report.duplicatesMerged} 条`) if ( report.retiredRemoved === 0 && report.zeroSuccessRetired === 0 && report.duplicatesMerged === 0 ) { lines.push('✅ 免疫记忆干净，无需清理') } return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // CRSI Critique — Self-Critique Hook (RLAIF) // ═══════════════════════════════════════════════════════════════ const crsiCritiqueCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const engine = ctx.engine const sc = engine.getSelfCritique?.() if (!sc) { return { content: t('commands.crsi_critique.not_init') } } const sub = args[0]?.toLowerCase() const config = sc.getConfig() if (sub === 'on' || sub === 'enable') { sc.setEnabled(true) return { content: [ t('commands.crsi_critique.on_title'), '', t('commands.crsi_critique.model', { model: config.model || 'auto (fastest available)' }), t('commands.crsi_critique.threshold', { pct: (config.threshold * 100).toFixed(0) }), t('commands.crsi_critique.target_tools', { tools: config.targetTools.join(', ') }), '', t('commands.crsi_critique.on_desc1'), t('commands.crsi_critique.on_desc2'), '', t('commands.crsi_critique.on_hint'), ].join('\n'), } } if (sub === 'off' || sub === 'disable') { sc.setEnabled(false) return { content: t('commands.crsi_critique.off_title') } } // status (default) const lines: string[] = [ t('commands.crsi_critique.status_title'), '', t('commands.crsi_critique.state', { state: config.enabled ? '🟢 Enabled' : '⚫ Disabled', }), t('commands.crsi_critique.model', { model: config.model || 'auto (fastest available)' }), t('commands.crsi_critique.threshold_detail', { pct: (config.threshold * 100).toFixed(0) }), t('commands.crsi_critique.target_tools', { tools: config.targetTools.length > 0 ? config.targetTools.join(', ') : 'all', }), t('commands.crsi_critique.timeout', { ms: String(config.timeoutMs) }), '', '──', t('commands.crsi_critique.status_on_hint'), t('commands.crsi_critique.status_off_hint'), '', t('commands.crsi_critique.inspired'), ] return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // CRSI Interpret — Tool-Call Behavior Dashboard // ═══════════════════════════════════════════════════════════════ const crsiInterpretCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const engine = ctx.engine const toolFilter = args[0]?.toLowerCase() const lines: string[] = [t('commands.crsi_interpret.title'), ''] // ── Error Signature Analysis ── const db = engine.getErrorSignatureDB?.() if (db) { const sigs = toolFilter ? db.getActive().filter((s) => s.toolName.toLowerCase() === toolFilter) : db.getActive() if (sigs.length === 0) { lines.push( toolFilter ? t('commands.crsi_interpret.no_sigs_for_tool', { tool: toolFilter }) : t('commands.crsi_interpret.no_sigs'), ) } else { lines.push( t('commands.crsi_interpret.sigs_title', { tool: toolFilter ? ` — \`${toolFilter}\`` : '', }), '', ) for (const sig of sigs.slice(0, 15)) { const bar = '█'.repeat(Math.round(sig.successRate * 10)) + '░'.repeat(10 - Math.round(sig.successRate * 10)) lines.push(`- **${sig.toolName}**: \`${sig.pattern.slice(0, 60)}\``) lines.push( t('commands.crsi_interpret.sig_success', { bar, pct: String(Math.round(sig.successRate * 100)), occurrences: String(sig.occurrences), strategy: sig.fixStrategy, }), ) lines.push(t('commands.crsi_interpret.sig_fix', { fix: sig.fixAction.slice(0, 80) })) lines.push('') } if (sigs.length > 15) lines.push(t('commands.crsi_interpret.sigs_more', { count: String(sigs.length - 15) }), '') } } // ── CRSI Reflection Summary ── const autoMemory = engine.getAutoMemory?.() if (autoMemory) { const count = autoMemory.sessionReflectionCount ?? 0 if (count > 0) { lines.push(t('commands.crsi_interpret.reflection_title'), '') lines.push(t('commands.crsi_interpret.reflections', { count: String(count) })) lines.push('') } } // ── Usage Tracker by Tool ── const usage = engine.getUsageTracker?.() if (usage) { const summary = usage.getSummary() lines.push(t('commands.crsi_interpret.usage_title'), '') lines.push( t('commands.crsi_interpret.api_in', { count: summary.apiInputTokens.toLocaleString() }), ) lines.push( t('commands.crsi_interpret.api_out', { count: summary.apiOutputTokens.toLocaleString() }), ) lines.push( t('commands.crsi_interpret.est_tokens', { count: summary.estimatedTokens.toLocaleString() }), ) lines.push('') } // ── System Health ── const meta = engine.getMetaRuleEngine?.() if (meta) { try { const analysis = meta.analyze() if (analysis.systemHealth) { const h = analysis.systemHealth const bar = '█'.repeat(Math.round(h.score / 10)) + '░'.repeat(10 - Math.round(h.score / 10)) lines.push(t('commands.crsi_interpret.health_title'), '') lines.push(t('commands.crsi_interpret.overall', { bar, score: String(h.score) })) lines.push(t('commands.crsi_interpret.assessment', { assessment: h.assessment })) if (h.components) { lines.push('') for (const [comp, score] of Object.entries(h.components)) { const cbar = '▮'.repeat(Math.round((score as number) / 10)) + '▯'.repeat(10 - Math.round((score as number) / 10)) lines.push(` ${comp.padEnd(20)} ${cbar} ${score}/100`) } } lines.push('') } } catch { // Meta analysis unavailable } } // ── Constitution Health ── const constitution = engine.getConstitutionLoader?.() if (constitution) { const c = constitution.load() const blocks = c.principles.filter((p) => p.enforce === 'block').length const warns = c.principles.filter((p) => p.enforce === 'warn').length const autos = c.principles.filter((p) => p.enforce === 'auto').length lines.push(t('commands.crsi_interpret.constitution_title'), '') lines.push( t('commands.crsi_interpret.principles', { total: String(c.principles.length), block: String(blocks), warn: String(warns), auto: String(autos), }), ) lines.push(t('commands.crsi_interpret.version', { version: c.version })) lines.push('') } if (lines.length <= 2) { lines.push(t('commands.crsi_interpret.empty')) } lines.push('──') lines.push(t('commands.crsi_interpret.filter_hint')) return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // CRSI Red-Team — Adversarial Self-Testing // ═══════════════════════════════════════════════════════════════ const crsiRedTeamCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const engine = ctx.engine const constitution = engine.getConstitutionLoader?.() const preflight = engine.getPreFlightChecker?.() if (!constitution || !preflight) { return { content: t('commands.crsi_red_team.not_init') } } const { RedTeam: RT } = await import('../../src/core/red-team.js') const redTeam = new RT() const report = redTeam.run(constitution, preflight) const lines: string[] = [ t('commands.crsi_red_team.title'), '', t('commands.crsi_red_team.score', { score: String(report.score) }), '', t('commands.crsi_red_team.metric'), t('commands.crsi_red_team.metric_sep'), t('commands.crsi_red_team.total_attacks', { count: String(report.total) }), t('commands.crsi_red_team.blocked', { count: String(report.blocked) }), t('commands.crsi_red_team.passed_through', { count: String(report.passedThrough) }), t('commands.crsi_red_team.false_positives', { count: String(report.falsePositives) }), '', ] // Per-principle breakdown lines.push(t('commands.crsi_red_team.by_principle'), '') for (const [pid, stats] of Object.entries(report.byPrinciple)) { const pct = stats.total > 0 ? Math.round((stats.blocked / stats.total) * 100) : 100 const bar = '█'.repeat(Math.round(pct / 10)) + '░'.repeat(10 - Math.round(pct / 10)) const icon = pct === 100 ? '✅' : pct >= 80 ? '⚠️' : '🔴' lines.push(`- ${icon} **${pid}**: ${bar} ${pct}% (${stats.blocked}/${stats.total})`) } lines.push('') // Detail: passed-through attacks const gaps = report.results.filter((r) => r.attack.shouldBlock && !r.blocked) if (gaps.length > 0) { lines.push(t('commands.crsi_red_team.gaps_title'), '') for (const g of gaps) { lines.push(`- **${g.attack.principleId}**: ${g.attack.description}`) lines.push( t('commands.crsi_red_team.gap_tool', { tool: g.attack.toolName, params: JSON.stringify(g.attack.params).slice(0, 80), }), ) } lines.push('') } // Detail: caught attacks const caught = report.results.filter((r) => r.blocked) if (caught.length > 0) { lines.push(t('commands.crsi_red_team.blocked_title'), '') for (const c of caught) { lines.push( t('commands.crsi_red_team.caught_by', { principle: c.attack.principleId, desc: c.attack.description, caught: c.caughtBy ?? 'unknown', }), ) } lines.push('') } if (report.score === 100) { lines.push(t('commands.crsi_red_team.all_blocked')) } else if (report.score >= 80) { lines.push(t('commands.crsi_red_team.good_coverage')) } else { lines.push(t('commands.crsi_red_team.critical')) } return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Auto-Dream: Background Memory Consolidation // ═══════════════════════════════════════════════════════════════ const dreamCmd: CommandHandler = async (ctx, args) => { const t = resolveT(ctx) const aggressive = args.includes('--aggressive') || args.includes('-a') const status = args.includes('--status') || args.includes('-s') const engine = ctx.engine.getDreamEngine?.() if (!engine) { return { content: t('commands.dream.not_init') } } if (status) { const history = engine.getDreamHistory() if (history.length === 0) { return { content: t('commands.dream.no_history') } } const lines: string[] = [t('commands.dream.history_title'), ''] for (const entry of history.slice(0, 5)) { const ts = new Date(entry.timestamp).toLocaleString('zh-CN') const total = entry.actions.length const applied = entry.actions.filter((a) => a.autoApplied).length const flagged = entry.actions.filter((a) => !a.autoApplied).length lines.push( t('commands.dream.history_entry', { time: ts, total: String(total), applied: String(applied), flagged: String(flagged), }), ) } return { content: lines.join('\n') } } const report = engine.dream({ aggressive }) const lines: string[] = [ t('commands.dream.done_title'), '', t('commands.dream.integrate', { before: String(report.beforeCount), after: String(report.afterCount), }), ...(report.phases.deduplicated > 0 ? [t('commands.dream.dedup', { count: String(report.phases.deduplicated) })] : []), ...(report.phases.contradictionsFound > 0 ? [t('commands.dream.contradiction', { count: String(report.phases.contradictionsFound) })] : []), ...(report.phases.merged > 0 ? [t('commands.dream.merge', { count: String(report.phases.merged) })] : []), ...(report.phases.solidified > 0 ? [t('commands.dream.solidify', { count: String(report.phases.solidified) })] : []), ...(report.phases.pruned > 0 ? [t('commands.dream.prune', { count: String(report.phases.pruned) })] : []), '', ] const flagged = report.actions.filter((a) => !a.autoApplied) if (flagged.length > 0 && !aggressive) { lines.push(t('commands.dream.review_hint', { count: String(flagged.length) })) lines.push('') for (const action of flagged) { lines.push(`- [ ] **${action.type}**: ${action.description}`) } } if (report.actions.length === 0) { lines.push(t('commands.dream.clean')) } return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Constitution // ═══════════════════════════════════════════════════════════════ const constitutionCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const engine = ctx.engine const constitution = engine.getConstitutionLoader?.() if (!constitution) { return { content: t('commands.constitution.not_init') } } const subCmd = args[0]?.toLowerCase() // /constitution reload if (subCmd === 'reload') { const c = constitution.reload() return { content: [ t('commands.constitution.reload_title'), '', t('commands.constitution.version', { version: c.version }), t('commands.constitution.principles', { count: String(c.principles.length) }), t('commands.constitution.path', { path: constitution.getPath() }), '', c.principles.map((p) => `- **${p.id}** [${p.enforce}]: ${p.text}`).join('\n'), ].join('\n'), } } // /constitution view (default) const c = constitution.load() const lines: string[] = [ t('commands.constitution.view_title'), '', t('commands.constitution.view_summary', { version: c.version, count: String(c.principles.length), path: constitution.getPath(), }), '', '---', '', ] for (const p of c.principles) { const icon = p.enforce === 'block' ? '🚫' : p.enforce === 'warn' ? '⚠️' : '🔄' lines.push(`### ${icon} ${p.id} [${p.enforce}]`) lines.push('') lines.push(p.text) if (p.rationale) lines.push(` *${p.rationale}*`) if (p.scope) lines.push(t('commands.constitution.scope', { scope: p.scope })) if (p.tools) lines.push(t('commands.constitution.tools', { tools: p.tools.join(', ') })) lines.push('') } lines.push('---') lines.push('') lines.push(t('commands.constitution.reload_hint')) lines.push(t('commands.constitution.edit_hint')) lines.push(t('commands.constitution.reset_hint')) lines.push('') lines.push(t('commands.constitution.inspired')) return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Bug Report // ═══════════════════════════════════════════════════════════════ const bugReportCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const engine = ctx.engine const lines: string[] = [ t('commands.bug_report.title'), '', t('commands.bug_report.hint'), '', '---', '', ] // ── Version & Runtime ── lines.push(t('commands.bug_report.env_title'), '') lines.push(t('commands.bug_report.mipham', { version: ctx.version })) lines.push(t('commands.bug_report.session', { id: ctx.sessionId })) lines.push(t('commands.bug_report.node', { version: process.version })) lines.push(t('commands.bug_report.platform', { platform: process.platform, arch: process.arch })) lines.push( t('commands.bug_report.os', { os: process.platform === 'darwin' ? 'macOS' : process.platform === 'linux' ? 'Linux' : 'Windows', }), ) lines.push(t('commands.bug_report.provider', { provider: ctx.providerId })) lines.push(t('commands.bug_report.model', { model: ctx.modelId })) lines.push('') // ── SIS errors ── try { const db = engine.getErrorSignatureDB?.() if (db) { const active = db.getActive() if (active.length > 0) { lines.push(t('commands.bug_report.sigs_title'), '') for (const sig of active.slice(0, 10)) { lines.push( t('commands.bug_report.sig_entry', { id: sig.id, pattern: sig.pattern, category: sig.category, occurrences: String(sig.occurrences), pct: (sig.successRate * 100).toFixed(0), }), ) } lines.push('') } } } catch { // SIS unavailable } // ── Hook health ── try { const hookEngine = engine.getHookEngine?.() if (hookEngine) { const health = hookEngine.getHookHealth() const disabled = health.filter((h) => h.health.disabled) if (disabled.length > 0) { lines.push(t('commands.bug_report.disabled_hooks'), '') for (const entry of disabled) { lines.push( t('commands.bug_report.hook_entry', { key: entry.key, failures: String(entry.health.failures), time: new Date(entry.health.disabledAt).toISOString(), }), ) } lines.push('') } } } catch { // Hook health unavailable } // ── Recent dream log ── try { const dreamEngine = engine.getDreamEngine?.() if (dreamEngine) { const history = dreamEngine.getDreamHistory() if (history.length > 0) { const last = history[0]! lines.push(t('commands.bug_report.last_dream'), '') lines.push(t('commands.bug_report.ran_at', { time: last.timestamp })) lines.push( t('commands.bug_report.actions', { total: String(last.actions.length), applied: String( last.actions.filter((a: { autoApplied: boolean }) => a.autoApplied).length, ), }), ) lines.push('') } } } catch { // Dream engine unavailable } // ── Steps to reproduce (template) ── lines.push(t('commands.bug_report.steps_title'), '') lines.push('1. ') lines.push('2. ') lines.push('3. ') lines.push('') lines.push(t('commands.bug_report.expected'), '') lines.push('') lines.push(t('commands.bug_report.actual'), '') lines.push('') lines.push('---') lines.push(t('commands.bug_report.generated', { version: ctx.version })) return { content: lines.join('\n'), copyContent: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Feedback // ═══════════════════════════════════════════════════════════════ // Changelog // ═══════════════════════════════════════════════════════════════ const changelogCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const lines: string[] = [t('commands.changelog.title'), ''] try { const { execSync } = await import('node:child_process') const tags = execSync('git tag --sort=-creatordate', { encoding: 'utf-8', timeout: 5000 }) .trim() .split('\n') .filter(Boolean) .slice(0, 15) if (tags.length === 0) { lines.push(t('commands.changelog.no_tags')) return { content: lines.join('\n') } } lines.push(t('commands.changelog.current', { version: ctx.version }), '') for (let i = 0; i < tags.length; i++) { const tag = tags[i]! const nextTag = tags[i + 1] try { const range = nextTag ? `${nextTag}..${tag}` : tag const log = execSync( `git log --oneline --no-merges ${range} --format="%s" 2>/dev/null | head -8`, { encoding: 'utf-8', timeout: 5000 }, ) .trim() .split('\n') .filter(Boolean) const isCurrent = tag === `v${ctx.version}` lines.push(`### ${tag}${isCurrent ? t('commands.changelog.current_marker') : ''}`, '') if (log.length > 0) { for (const entry of log) { lines.push(`- ${entry}`) } } else { lines.push(t('commands.changelog.release_only')) } lines.push('') } catch { lines.push(`### ${tag}`, '', t('commands.changelog.details_unavailable'), '') } } } catch { lines.push(t('commands.changelog.no_git')) } lines.push('---') lines.push(t('commands.changelog.full_history')) return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Workflow // ═══════════════════════════════════════════════════════════════ const planCmd: CommandHandler = (_ctx, args) => { const t = resolveT(_ctx) const description = args.join(' ') || undefined return { content: t('commands.plan.content'), forwardToAI: description ? `Use EnterPlanMode with description: "${description}". Then explore, design, and use ExitPlanMode when ready for approval.` : 'Use EnterPlanMode to enter plan mode. Explore the codebase, design an approach, then use ExitPlanMode to submit for approval.', } } const tddCmd: CommandHandler = (_ctx, args) => { const t = resolveT(_ctx) const target = args.join(' ') || 'the current task' return { content: stripIndent` ${t('commands.tdd.title')} ${t('commands.tdd.start', { target })} ${t('commands.tdd.cycle')} ${t('commands.tdd.red')} ${t('commands.tdd.green')} ${t('commands.tdd.refactor')} ${t('commands.tdd.guide')} `, forwardToAI: `Follow the Test-Driven Development workflow for ${target}: 1. RED — Write a failing test that defines the expected behavior. Use the project's test framework (Vitest/Jest/pytest). Show me the test code and confirm it fails. 2. GREEN — Write the minimum code needed to make the test pass. Run the test to verify it passes. 3. REFACTOR — Clean up both test and implementation code. Remove duplication, improve names, simplify. Keep tests green. Repeat for each behavior. Do NOT write implementation before tests.`, } } const todosCmd: CommandHandler = (_ctx, args) => { const t = resolveT(_ctx) const sub = args[0] if (sub === 'create') { const title = args.slice(1).join(' ') if (!title.trim()) { return { content: t('commands.todos.usage_create') } } return { content: `${t('commands.todos.create_title')}\n\nCreating task: "${title.trim()}"\n\nPassing to AI for structured task creation with the Task tool...`, forwardToAI: `Create a new task using the Task tool with action "create" and subject "${title.trim()}". Set a clear description and activeForm.`, } } if (sub === 'list' || !sub) { return { content: stripIndent` ${t('commands.todos.list_title')} ${t('commands.todos.list_fetching')} Shortcuts: ${t('commands.todos.item_list')} ${t('commands.todos.item_create')} `, forwardToAI: 'Use the Task tool with action "list" to show all current tasks. Present them in a clear summary grouped by status (pending/in_progress/completed). If there are no tasks, suggest creating one.', } } return { content: stripIndent` ${t('commands.todos.default_title')} ${t('commands.todos.item_list')} ${t('commands.todos.item_create')} ${t('commands.todos.default_body')} `, forwardToAI: 'Use the Task tool with action "list" to show all current tasks, then present them clearly.', } } // ═══════════════════════════════════════════════════════════════ // Project // ═══════════════════════════════════════════════════════════════ // ═══════════════════════════════════════════════════════════════ // Phase 1 — New Session Commands // ═══════════════════════════════════════════════════════════════ const renameCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const name = args.join(' ') if (!name.trim()) { return { content: t('commands.rename.usage') } } ctx.setSessionTitle(name.trim()) // Persist to the cross-session registry so `@name` can address this session. // Returns the (possibly uniquified) final name — surface that to the user. const finalName = renameActiveSession(ctx.sessionId, name.trim()) return { content: t('commands.rename.confirmed', { name: finalName ?? name.trim() }) } } const goalCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const input = args.join(' ') // ── Show goal status ── if (!input.trim()) { const state = ctx.engine.getGoalState?.() || { goal: undefined, decompose: false, subtasks: [] } if (!state.goal) { return { content: t('commands.goal.usage') } } const lines = [ t('commands.goal.status_title'), '', t('commands.goal.status_goal', { goal: state.goal }), ] if (state.verifyScript) lines.push(t('commands.goal.status_verify_script', { script: state.verifyScript })) if (state.verifySkill) lines.push(t('commands.goal.status_verify_skill', { skill: state.verifySkill })) if (state.decompose) lines.push(t('commands.goal.status_decompose', { count: String(state.subtasks.length) })) lines.push('', t('commands.goal.status_clear_hint')) return { content: lines.join('\n') } } // ── Parse options ── let goal = input let decompose = false let verifyScript: string | undefined let verifySkill: string | undefined // Extract --decompose if (goal.includes(' --decompose')) { decompose = true goal = goal.replace(' --decompose', '') } // Extract --verify-script const scriptMatch = goal.match(/ --verify-script\s+(\S+)/) if (scriptMatch) { verifyScript = scriptMatch[1]! goal = goal.replace(` --verify-script ${verifyScript}`, '') } // Extract --verify-skill const skillMatch = goal.match(/ --verify-skill\s+(\S+)/) if (skillMatch) { verifySkill = skillMatch[1]! goal = goal.replace(` --verify-skill ${verifySkill}`, '') } goal = goal.trim() if (!goal) { return { content: t('commands.goal.empty_set') } } ctx.setGoal(goal) ctx.engine.setGoal(goal, { verifyScript, verifySkill, decompose }) const lines = [t('commands.goal.set', { goal })] if (decompose) { lines.push(t('commands.goal.decompose_enabled')) // Decompose by creating initial subtasks const decomposeMsg = `Break down this goal into 3-5 subtasks: "${goal}". For each subtask, use the Task tool with action "create", giving the subject and description. Mark each as blocked by the previous one to create a dependency chain.` return { content: lines.join('\n'), forwardToAI: decomposeMsg, } } if (verifyScript) lines.push(t('commands.goal.status_verify_script', { script: verifyScript })) if (verifySkill) lines.push(t('commands.goal.status_verify_skill', { skill: verifySkill })) lines.push('') lines.push(t('commands.goal.status_hint')) return { content: lines.join('\n') } } const recapCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const msgs = c.getMessages() if (msgs.length === 0) { return { content: t('commands.recap.no_conversation') } } const userMsgs = msgs.filter((m) => m.role === 'user').length const assistantMsgs = msgs.filter((m) => m.role === 'assistant').length const tokens = c.getEstimatedTokens() const checkpointCount = c.getCheckpoints().length const topics = msgs .filter((m) => m.role === 'user' && typeof m.content === 'string') .slice(0, 5) .map((m) => { const text = typeof m.content === 'string' ? m.content : '' return text.length > 80 ? text.slice(0, 80) + '...' : text }) return { content: stripIndent` ${t('commands.recap.title')} ${t('commands.recap.messages')} ${msgs.length} (${userMsgs} user, ${assistantMsgs} assistant) ${t('commands.recap.est_tokens')} ~${tokens.toLocaleString()} ${t('commands.recap.checkpoints')} ${checkpointCount} ${t('commands.recap.recent_topics')} ${topics.map((t, i) => ` ${i + 1}. ${t}`).join('\n')} ${t('commands.recap.footer')} `, } } const usageCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const estTokens = c.getEstimatedTokens() const msgs = c.getMessages() const maxTokens = c.getMaxTokens() const pct = ((estTokens / maxTokens) * 100).toFixed(1) const tracker = ctx.engine.getUsageTracker() const summary = tracker.getSummary() // Build per-tool breakdown const toolLines: string[] = [] const sortedTools = Object.entries(summary.tools).sort( (a, b) => b[1].inputTokens + b[1].outputTokens - (a[1].inputTokens + a[1].outputTokens), ) for (const [name, usage] of sortedTools) { const total = usage.inputTokens + usage.outputTokens const prefix = name.startsWith('mcp__') ? '🔌 ' : ' ' toolLines.push( `${prefix}${name.padEnd(18)} ${total.toLocaleString().padStart(8)} tokens (${usage.calls} call${usage.calls !== 1 ? 's' : ''})`, ) } const apiTotal = summary.apiInputTokens + summary.apiOutputTokens const apiLine = summary.apiInputTokens > 0 || summary.apiOutputTokens > 0 ? t('commands.usage.api_tokens', { in: summary.apiInputTokens.toLocaleString(), out: summary.apiOutputTokens.toLocaleString(), total: apiTotal.toLocaleString(), }) : t('commands.usage.no_api') const toolSection = toolLines.length > 0 ? `\n${t('commands.usage.tool_section')}\n${toolLines.join('\n')}` : '' const loops = listActiveAutoloops() const loopSection = loops.length > 0 ? `\n${t('commands.usage.loops_header')}\n${formatLoopRows(loops).join('\n')}` : '' return { content: stripIndent` ${t('commands.usage.title')} ${apiLine} ${t('commands.usage.context_line', { est: estTokens.toLocaleString(), max: maxTokens.toLocaleString(), pct })} ${t('commands.usage.messages_line', { count: String(msgs.length) })} ${t('commands.usage.provider_line', { provider: ctx.providerId })} ${t('commands.usage.model_line', { model: ctx.modelId })} ${'█'.repeat(Math.ceil(Number(pct) / 5))}${'░'.repeat(20 - Math.ceil(Number(pct) / 5))} ${pct}% ${toolSection}${loopSection} ${t('commands.usage.hint')} `, } } const reloadSkillsCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) if (!ctx.skillsLoader) { return { content: t('commands.reload_skills.unavailable') } } try { const config = ctx.config ctx.skillsLoader.loadBuiltin(process.cwd()) if (config.skills?.paths) { ctx.skillsLoader.loadExternal(config.skills.paths) } const skills = ctx.skillsLoader.list() return { content: `${t('commands.reload_skills.confirmed', { count: String(skills.length) })}\n\n${skills.map((s) => ` ${s.name.padEnd(28)} ${s.type.padEnd(10)} ${s.description}`).join('\n')}`, } } catch (err) { return { content: t('commands.reload_skills.failed', { error: String(err) }) } } } // ═══════════════════════════════════════════════════════════════ // Skill Marketplace — Community skill registry + installation // ═══════════════════════════════════════════════════════════════ const browseSkillsCmd: CommandHandler = async (ctx) => { const _t = resolveT(ctx) const { getAvailableSkills, listInstalledSkills } = await import('../skills/registry') const available = getAvailableSkills() const installed = new Set( listInstalledSkills().map((f: string) => f.replace(/\.(SKILL\.)?md$/i, '')), ) const categories = new Map() for (const s of available) { const list = categories.get(s.category) || [] list.push(s.name) categories.set(s.category, list) } const lines: string[] = [ '── Community Skills ──', '', `${available.length} skills available · ${installed.size} installed`, '', ] for (const [cat, names] of categories) { lines.push(` ${cat}:`) for (const name of names) { const entry = available.find((s) => s.name === name)! const marker = entry.builtin ? '🏠' : installed.has(name) ? '✅' : '⬜' const status = entry.builtin ? ' (built-in)' : installed.has(name) ? ' (installed)' : '' lines.push(` ${marker} /${name.padEnd(26)} ${entry.description}${status}`) } lines.push('') } lines.push('Install: /install-skill ') lines.push('Install from URL: /install-skill ') lines.push('Remove: /remove-skill ') return { content: lines.join('\n') } } const installSkillCmd: CommandHandler = async (ctx, args) => { const { installSkill, installSkillFromUrl } = await import('../skills/registry') const target = args[0] if (!target) { return { content: 'Usage: /install-skill or /install-skill ' } } const marketplaceConfig = ctx.config.marketplace let result: { success: boolean; name: string; message: string } if (target.startsWith('http://') || target.startsWith('https://')) { result = await installSkillFromUrl(target, marketplaceConfig) } else { result = await installSkill(target, marketplaceConfig) } return { content: result.success ? `✅ ${result.message}` : `❌ ${result.message}`, } } const removeSkillCmd: CommandHandler = async (_ctx, args) => { const { removeSkill } = await import('../skills/registry') const name = args[0] if (!name) { return { content: 'Usage: /remove-skill ' } } const result = removeSkill(name) return { content: result.success ? `✅ ${result.message}` : `❌ ${result.message}`, } } const marketplaceCmd: CommandHandler = async (_ctx, args) => { const { readMarketplaces, addMarketplace, removeMarketplace, isValidMarketplaceRef, MARKETPLACES_PATH, } = await import('../skills/marketplace') const { readFileSync, existsSync, mkdirSync } = await import('node:fs') const { atomicWriteFileSync } = await import('../shared/atomic-write') const { dirname } = await import('node:path') const readSources = () => readMarketplaces((p) => (existsSync(p) ? readFileSync(p, 'utf-8') : null)) const sub = args[0] const ref = args[1] if (!sub || sub === 'list') { const sources = readSources() const lines = ['── Marketplace Sources ──', ''] for (const s of sources) lines.push(` ${s.owner}/${s.repo}`) lines.push( '', 'Add: /marketplace add /', 'Remove: /marketplace remove /', ) return { content: lines.join('\n') } } if (sub === 'add' || sub === 'remove') { if (!ref) return { content: `Usage: /marketplace ${sub} /` } const [owner, repo] = ref.split('/') if (!owner || !repo || !isValidMarketplaceRef(owner, repo)) { return { content: `❌ Invalid marketplace ref "${ref}". Expected /.` } } const current = readSources() const result = sub === 'add' ? addMarketplace(current, owner, repo) : removeMarketplace(current, owner, repo) const changed = 'added' in result ? result.added : result.removed try { mkdirSync(dirname(MARKETPLACES_PATH), { recursive: true }) atomicWriteFileSync(MARKETPLACES_PATH, JSON.stringify(result.sources, null, 2), { mode: 0o644, }) } catch (err) { return { content: `❌ Failed to save marketplaces: ${String(err)}` } } if (sub === 'add') { return { content: changed ? `✅ Added ${owner}/${repo}` : `ℹ ${owner}/${repo} already exists`, } } return { content: changed ? `✅ Removed ${owner}/${repo}` : `ℹ ${owner}/${repo} not found` } } return { content: 'Usage: /marketplace add|list|remove [owner/repo]' } } const browseMarketplaceCmd: CommandHandler = async (_ctx) => { const { readMarketplaces, discoverSkills, downloadFile } = await import('../skills/marketplace') const { readFileSync, existsSync } = await import('node:fs') const sources = readMarketplaces((p) => (existsSync(p) ? readFileSync(p, 'utf-8') : null)) const lines = ['── Marketplace Skills ──', ''] for (const source of sources) { lines.push(` ${source.owner}/${source.repo}:`) try { const skills = await discoverSkills(source, globalThis.fetch, downloadFile) if (skills.length === 0) lines.push(' (no skills found)') for (const s of skills) lines.push(` ${s.name.padEnd(24)} ${s.description}`) } catch { lines.push(' (failed to discover — check network / rate limit)') } lines.push('') } lines.push('Install: /install-skill ') return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Plugins — Community plugin marketplace + local management // ═══════════════════════════════════════════════════════════════ const pluginsCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const manager = ctx.pluginManager if (!manager) { return { content: t('commands.plugins.unavailable') } } const plugins = manager.list() if (plugins.length === 0) { return { content: t('commands.plugins.none') } } const lines: string[] = [ t('commands.plugins.title'), '', ...plugins.map((p) => { const status = p.enabled ? '✅ enabled' : '⛔ disabled' return ` ${p.name.padEnd(24)} v${p.version.padEnd(10)} ${status} ${p.installedAt ? new Date(p.installedAt).toLocaleDateString() : ''}` }), '', `${plugins.length} plugin(s) installed.`, '', 'Manage: /install-plugin | /remove-plugin | /plugin-enable | /plugin-disable', ] return { content: lines.join('\n') } } const browsePluginsCmd: CommandHandler = async (ctx) => { const { getAvailablePlugins } = await import('../plugin/plugin-registry') const available = getAvailablePlugins() const installed = new Set((ctx.pluginManager?.list() ?? []).map((p) => p.name)) const categories = new Map() for (const p of available) { const list = categories.get(p.category) || [] list.push(p.name) categories.set(p.category, list) } const lines: string[] = [ '── Community Plugins ──', '', `${available.length} plugins available · ${installed.size} installed`, '', ] for (const [cat, names] of categories) { lines.push(` ${cat}:`) for (const name of names) { const entry = available.find((p) => p.name === name)! const marker = installed.has(name) ? '✅' : '⬜' lines.push(` ${marker} ${name.padEnd(22)} ${entry.description}`) lines.push(` npm: ${entry.npmPackage}`) } lines.push('') } lines.push('Install: /install-plugin ') lines.push('Remove: /remove-plugin ') return { content: lines.join('\n') } } const installPluginCmd: CommandHandler = async (ctx, args) => { const target = args[0] if (!target) { return { content: [ 'Usage: /install-plugin or /install-plugin ', '', 'Examples:', ' /install-plugin @roomi-fields/notebooklm-mcp', ' /install-plugin mipham-plugin-security', ' /install-plugin ~/my-custom-plugin/', '', 'Use /browse-plugins to see available community plugins.', ].join('\n'), } } const manager = ctx.pluginManager if (!manager) { return { content: '❌ PluginManager not available in this session.' } } // Determine if it's a local path or npm package const isLocalPath = target.startsWith('./') || target.startsWith('/') || target.startsWith('~/') let result: { success: boolean; message: string } if (isLocalPath) { const { resolve } = await import('node:path') const resolved = resolve(target.replace(/^~/, homedir())) result = manager.install(resolved) } else { // Treat as npm package name result = manager.installFromNpm(target) } return { content: result.success ? `✅ ${result.message}` : `❌ ${result.message}`, } } const removePluginCmd: CommandHandler = (ctx, args) => { const name = args[0] if (!name) { return { content: 'Usage: /remove-plugin \n\nUse /plugins to see installed plugins.' } } const manager = ctx.pluginManager if (!manager) { return { content: '❌ PluginManager not available in this session.' } } const ok = manager.remove(name) return { content: ok ? `✅ Plugin "${name}" removed.` : `❌ Plugin "${name}" not found. Use /plugins to list installed plugins.`, } } const pluginEnableCmd: CommandHandler = (ctx, args) => { const name = args[0] if (!name) { return { content: 'Usage: /plugin-enable \n\nUse /plugins to see installed plugins.' } } const manager = ctx.pluginManager if (!manager) { return { content: '❌ PluginManager not available in this session.' } } const ok = manager.enable(name) return { content: ok ? `✅ Plugin "${name}" enabled.` : `❌ Plugin "${name}" not found. Use /plugins to list installed plugins.`, } } const pluginDisableCmd: CommandHandler = (ctx, args) => { const name = args[0] if (!name) { return { content: 'Usage: /plugin-disable \n\nUse /plugins to see installed plugins.' } } const manager = ctx.pluginManager if (!manager) { return { content: '❌ PluginManager not available in this session.' } } const ok = manager.disable(name) return { content: ok ? `✅ Plugin "${name}" disabled.` : `❌ Plugin "${name}" not found. Use /plugins to list installed plugins.`, } } // ═══════════════════════════════════════════════════════════════ // Phase 1 — History & Checkpoint Commands // ═══════════════════════════════════════════════════════════════ const rewindCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const checkpoints = c.getCheckpoints() if (checkpoints.length === 0) { return { content: t('commands.rewind.no_checkpoints') } } const result = c.restoreCheckpoint() if (!result.restored) { return { content: t('commands.rewind.no_restore') } } return { content: t('commands.rewind.confirmed', { label: result.label, count: String(result.messageCount), remaining: String(c.getCheckpoints().length), }), clearMessages: true, } } const undoCmd: CommandHandler = rewindCmd const copyCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const msgs = c.getMessages() const assistantMsgs = msgs.filter((m) => m.role === 'assistant') if (assistantMsgs.length === 0) { return { content: t('commands.copy.no_responses') } } // Determine which response to copy: last N or last 1 let n = 1 if (args[0]) { n = parseInt(args[0]!, 10) if (isNaN(n) || n < 1) { return { content: t('commands.copy.usage') } } } const toCopy = assistantMsgs.slice(-n) const text = toCopy .map((m) => (typeof m.content === 'string' ? m.content : JSON.stringify(m.content))) .join('\n\n---\n\n') return { content: t('commands.copy.confirmed', { count: String(toCopy.length) }), copyContent: text, } } const diffCmd: CommandHandler = async (_ctx) => { const t = resolveT(_ctx) try { const { execSync } = await import('node:child_process') const output = execSync('git diff --stat', { encoding: 'utf-8', timeout: 5000 }) if (!output.trim()) { return { content: t('commands.diff.clean') } } // Get full diff but limit to reasonable size const fullDiff = execSync('git diff --no-color', { encoding: 'utf-8', timeout: 5000 }) const MAX_LINES = 60 const lines = fullDiff.split('\n') const truncated = lines.length > MAX_LINES ? lines.slice(0, MAX_LINES).join('\n') + `\n\n... (${lines.length - MAX_LINES} more lines. Use git diff to see full output.)` : fullDiff return { content: `${t('commands.diff.title')}\n\n${truncated}`, } } catch { return { content: t('commands.diff.error'), } } } // ═══════════════════════════════════════════════════════════════ // Phase 1 — Model Control Commands // ═══════════════════════════════════════════════════════════════ const fastCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const arg = args[0]?.toLowerCase() if (arg === 'on') { ctx.setFastMode(true) return { content: t('commands.fast.on') } } else if (arg === 'off') { ctx.setFastMode(false) return { content: t('commands.fast.off') } } else if (arg) { return { content: t('commands.fast.usage') } } else { return { content: t('commands.fast.unknown') } } } const effortCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const VALID_LEVELS = ['low', 'medium', 'high', 'xhigh', 'max'] const level = args[0]?.toLowerCase() if (!level || !VALID_LEVELS.includes(level)) { return { content: t('commands.effort.usage', { model: ctx.modelId }) } } ctx.setEffort(level) return { content: t('commands.effort.confirmed', { level }) } } const focusCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) ctx.setFocusMode(true) return { content: t('commands.focus.on') } } // ═══════════════════════════════════════════════════════════════ // Ultracode Mode — toggle multi-agent workflow orchestration // ═══════════════════════════════════════════════════════════════ const ultracodeCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const arg = args[0]?.toLowerCase() if (arg === 'on') { ctx.setUltracodeMode(true) return { content: t('commands.ultracode.on'), forwardToAI: stripIndent` Ultracode mode is now ACTIVE. For every substantive task: - Use the Workflow tool to orchestrate multi-agent execution - Default to pipeline() for multi-stage work — stages flow independently - Fan out independent work across parallel agents - Apply adversarial verification (verify with skeptics) to findings before accepting them - Synthesize results with a top-tier model pass - Prefer pipeline() over parallel() barriers — don't block stages unnecessarily - Use loopUntilConvergence when discovering unknown-size sets (bugs, issues, edge cases) - EDGE LOGIC IS FREE: flatten, dedupe, filter in plain JavaScript — NOT agent calls Keep responses concise — the workflow graph handles the depth.`, } } else if (arg === 'off') { ctx.setUltracodeMode(false) return { content: t('commands.ultracode.off'), forwardToAI: 'Ultracode mode is now OFF. Revert to standard single-agent execution. Do NOT use Workflow tool unless explicitly asked.', } } return { content: t('commands.ultracode.usage') } } // ═══════════════════════════════════════════════════════════════ // Phase 1 — Workflow Commands // ═══════════════════════════════════════════════════════════════ const tasksCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const msgs = c.getMessages() // Scan for task-related tool uses in message history. // 任务工具只有一个 `Task`，动作走 action 参数 —— 过滤条件必须按真实工具名匹配， // 否则计数恒为 0，「已检测到 N 次任务操作」这条分支永远不可达。 const toolUses = msgs.flatMap((m) => { if (Array.isArray(m.content)) { return m.content.filter((b) => b.type === 'tool_use' && b.name === 'Task') } return [] }) return { content: stripIndent` ${t('commands.task_list.title')} ${toolUses.length > 0 ? t('commands.task_list.detected', { count: String(toolUses.length) }) : t('commands.task_list.no_tasks')} ${t('commands.task_list.reference')} Task(action: "create") — create a new task Task(action: "list") — list all tasks Task(action: "update") — update task status Task(action: "get") — get task details Task(action: "delete") — delete a task Task(action: "output") — get background task output Task(action: "stop") — stop a running task ${t('commands.task_list.legacy_hint')} `, } } const branchCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const name = args.join(' ') || `branch-${Date.now()}` const c = ctx.engine.getContext() const msgs = c.getMessages() if (msgs.length === 0) { return { content: t('commands.branch.no_conversation') } } const checkpointId = c.saveCheckpoint(name) return { content: t('commands.branch.details', { name, checkpoint: String(checkpointId), messages: String(msgs.length), }), } } /** Parse human-readable interval to seconds. Returns null if invalid. */ function parseInterval(raw: string): number | null { const m = raw.trim().match(/^(\d+)\s*(s|sec|m|min|h|hr|hour)$/i) if (!m) return null const n = parseInt(m[1]!, 10) const unit = m[2]!.toLowerCase() if (unit === 's' || unit === 'sec') return n if (unit === 'm' || unit === 'min') return n * 60 if (unit === 'h' || unit === 'hr' || unit === 'hour') return n * 3600 return null } const loopCmd: CommandHandler = async (ctx, args) => { const sub = args[0] // ── /loop init — scaffold project vault ── if (sub === 'init' || sub === 'scaffold') { const targetPath = args[1] || process.cwd() const { scaffoldLoopKit } = await import('../commands/loop-scaffold') const { created, skipped } = scaffoldLoopKit(targetPath) const { resolve } = await import('node:path') const resolved = resolve(targetPath.replace(/^~/, homedir())) const lines: string[] = ['── LoopKit Vault Created ──', '', `Location: ${resolved}`, ''] if (created.length > 0) { lines.push(`Created (${created.length}):`) for (const c of created.slice(0, 12)) { const rel = c.replace(resolved + '/', '') lines.push(` ✅ ${rel}`) } if (created.length > 12) { lines.push(` ... and ${created.length - 12} more`) } } if (skipped.length > 0) { lines.push('') lines.push(`Skipped (${skipped.length} — already exist):`) for (const s of skipped.slice(0, 5)) { const rel = s.replace(resolved + '/', '') lines.push(` ⏭ ${rel}`) } if (skipped.length > 5) { lines.push(` ... and ${skipped.length - 5} more`) } } lines.push('') lines.push('── Structure ──') lines.push('') lines.push(' .mipham/ Mipham Code config directory') lines.push(' ├── CLAUDE.md Project AI instructions') lines.push(' ├── settings.json Permission & hook settings') lines.push(' ├── hooks/ Lifecycle hook scripts (3)') lines.push(' ├── agents/ Custom sub-agents') lines.push(' └── skills/ 9 domain-sorted skill directories') lines.push('') lines.push(' .mcp.json MCP server configuration') lines.push(' MEMORY.md AI persistent memory') lines.push(' run.sh / install.sh Project scripts') lines.push('') lines.push('Edit .mipham/CLAUDE.md to customize AI behavior.') lines.push('Add skills to .mipham/skills// directories.') lines.push('Configure MCP servers in .mcp.json.') lines.push('') lines.push('/loop init to scaffold in a different location.') return { content: lines.join('\n') } } // ── /loop stop — stop autonomous loop ── if (sub === 'stop') { const { completeAutoloopJournal, listActiveAutoloops } = await import('../commands/autoloop-journal.js') const active = listActiveAutoloops() if (active.length === 0) { return { content: 'No active autonomous loops to stop.' } } for (const j of active) completeAutoloopJournal(j.sessionId, 'stopped') return { content: `Stopped ${active.length} autonomous loop(s).` } } // ── /loop status — show autonomous loop progress ── if (sub === 'status') { const { listActiveAutoloops, getAutoloopStatus } = await import('../commands/autoloop-journal.js') const active = listActiveAutoloops() if (active.length === 0) { return { content: 'No active autonomous loops.\n\nUse /loop auto to start one.' } } const lines = active.map((j) => getAutoloopStatus(j.sessionId)) return { content: lines.join('\n\n') } } // ── /loop auto — autonomous self-paced loop ── if (sub === 'auto') { const prompt = args.slice(1).join(' ') if (!prompt.trim()) { return { content: 'Usage: /loop auto \n\n' + 'Start an autonomous self-paced loop. The AI will:\n' + '1. Work on the task\n' + '2. Decide how long to wait before the next iteration\n' + '3. Use ScheduleWakeup to self-schedule\n' + '4. Use ScheduleWakeup(stop:true) when the goal is reached\n\n' + 'Use /loop status to view progress.\n' + 'Use /loop stop to stop the loop.', } } const sessionId = `autoloop-${Date.now()}` const { createAutoloopJournal } = await import('../commands/autoloop-journal.js') createAutoloopJournal(sessionId, prompt, ctx.engine.getUsageTracker().totalApiTokens) const autoPrompt = `## Autonomous Loop — ${sessionId}\n\n` + `Task: ${prompt}\n\n` + `You are in an autonomous loop. Each iteration:\n` + `1. Make progress on the task above\n` + `2. When you need to wait (e.g., for a deploy, CI, external event), call ScheduleWakeup with:\n` + ` - delaySeconds: your best estimate of how long to wait (60-3600)\n` + ` - reason: one sentence explaining why\n` + ` - prompt: include this full autonomous loop ID: ${sessionId}\n` + `3. When the task is COMPLETE, call ScheduleWakeup with stop:true\n\n` + `After each iteration, log progress by reading/writing the journal at ~/.mipham/autoloop/${sessionId}.json.\n` + `Use the autoloop-journal module: logAutoloopIteration("${sessionId}", "").` return { content: `── Autonomous Loop Started ──\n\n` + `ID: ${sessionId}\n` + `Prompt: "${prompt.slice(0, 100)}${prompt.length > 100 ? '...' : ''}"\n\n` + `The AI will self-pace. Use /loop status to view progress.\n` + `Use /loop stop to stop.`, forwardToAI: autoPrompt, } } // ── /loop — recurring prompt (or auto-detect autonomous) ── if (args.length < 2) { return { content: stripIndent` Usage: /loop Fixed-interval loop or: /loop auto Autonomous self-paced loop or: /loop stop Stop active autonomous loop or: /loop status Show autonomous loop progress or: /loop init [path] Scaffold project vault structure Fixed interval formats: 10s — 10 seconds 5m — 5 minutes 1h — 1 hour Examples: /loop 5m check the deploy status /loop auto monitor CI and fix failing tests /loop init Create vault in current directory `, } } const interval = args[0]! const prompt = args.slice(1).join(' ') const seconds = parseInterval(interval) // Auto-detect: if first arg is not a valid interval, assume autonomous mode if (seconds === null) { // Could be: /loop const fullPrompt = args.join(' ') if (fullPrompt.length > 0) { const sessionId = `autoloop-${Date.now()}` const { createAutoloopJournal } = await import('../commands/autoloop-journal.js') createAutoloopJournal(sessionId, fullPrompt, ctx.engine.getUsageTracker().totalApiTokens) const autoPrompt = `## Autonomous Loop — ${sessionId}\n\n` + `Task: ${fullPrompt}\n\n` + `You are in an autonomous loop (auto-detected — no interval specified). Each iteration:\n` + `1. Make progress on the task above\n` + `2. When you need to wait, call ScheduleWakeup with your chosen delaySeconds and this loop ID: ${sessionId}\n` + `3. When COMPLETE, call ScheduleWakeup with stop:true\n\n` + `Log progress using logAutoloopIteration("${sessionId}", "").` return { content: `── Autonomous Loop Started (auto-detected) ──\n\n` + `ID: ${sessionId}\n` + `Prompt: "${fullPrompt.slice(0, 100)}${fullPrompt.length > 100 ? '...' : ''}"\n\n` + `No interval specified — using autonomous self-paced mode.\n` + `Use /loop stop to stop.`, forwardToAI: autoPrompt, } } return { content: `── Invalid Interval ──\n\n"${interval}" is not a recognised interval.\n\nUse formats like: 10s, 5m, 1h, 30min, 2hr\nOr use /loop auto for autonomous mode.`, } } // Fixed-interval loop — directly invoke ScheduleWakeup try { const { scheduleWakeupTool } = await import('../tools/scheduling/schedule-wakeup.js') const result = await scheduleWakeupTool.execute( { delaySeconds: seconds, reason: `Loop: ${prompt.slice(0, 40)}`, prompt }, { cwd: process.cwd(), sessionId: 'loop-session', provider: '', model: '' }, ) if (!result.success) { return { content: `Loop failed: ${result.error || 'unknown error'}` } } // When scheduling notices are hidden, stay silent — the loop is already scheduled. if (ctx.config.showSchedulingNotices === false) { return { content: '' } } return { content: result.content } } catch (e) { return { content: `Loop failed: ${e instanceof Error ? e.message : String(e)}` } } } const scheduleCmd: CommandHandler = async (_ctx) => { try { const { cronListTool } = await import('../tools/scheduling/cron.js') const result = await cronListTool.execute( {}, { cwd: process.cwd(), sessionId: 'schedule-view', provider: '', model: '' }, ) return { content: result.content } } catch (e) { return { content: `Schedule check failed: ${e instanceof Error ? e.message : String(e)}` } } } // ═══════════════════════════════════════════════════════════════ // Diagnostic // ═══════════════════════════════════════════════════════════════ const doctorCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const lines: string[] = [ t('commands.doctor.title'), '', `Mipham Code v${ctx.version}`, `Runtime ${typeof Bun !== 'undefined' ? 'Bun ' + Bun.version : 'Node.js ' + process.version}`, `Platform ${process.platform} ${process.arch}`, `CWD ${process.cwd()}`, `PID ${process.pid}`, '', t('commands.doctor.config_section'), `Provider ${ctx.providerId} / ${ctx.modelId}`, `Permission ${ctx.config.permission}`, `Providers ${ctx.config.providers.length} configured (${ctx.config.providers.filter((p) => p.status !== 'upcoming').length} active)`, '', t('commands.doctor.session_section'), ] const c = ctx.engine.getContext() const msgs = c.getMessages() const tokens = c.getEstimatedTokens() const maxTokens = c.getMaxTokens() lines.push(`Messages ${msgs.length}`) lines.push( `Tokens ~${tokens.toLocaleString()} / ${maxTokens.toLocaleString()} (${((tokens / maxTokens) * 100).toFixed(1)}%)`, ) lines.push(`Checkpoints ${c.getCheckpoints().length}`) // Git info try { const { execSync } = await import('node:child_process') lines.push('') lines.push(t('commands.doctor.git_section')) const branch = execSync('git branch --show-current', { encoding: 'utf-8', timeout: 3000, }).trim() lines.push(`Branch ${branch || '(detached)'}`) const status = execSync('git status --porcelain', { encoding: 'utf-8', timeout: 3000 }) const changed = status.trim().split('\n').filter(Boolean).length lines.push(`Changed ${changed} file${changed !== 1 ? 's' : ''}`) const log = execSync('git log --oneline -3', { encoding: 'utf-8', timeout: 3000 }).trim() lines.push(`Last commits ${log.split('\n').length}`) lines.push('') lines.push( log .split('\n') .map((l, i) => ` ${i + 1}. ${l}`) .join('\n'), ) } catch { lines.push('') lines.push(t('commands.doctor.git_section')) lines.push(t('commands.doctor.no_git')) } // Skills info if (ctx.skillsLoader) { lines.push('') lines.push(t('commands.doctor.skills_section')) try { const skills = ctx.skillsLoader.list() const standard = skills.filter((s: { type: string }) => s.type === 'standard').length const mipham = skills.filter((s: { type: string }) => s.type === 'mipham').length lines.push(`Loaded ${skills.length} (${standard} standard + ${mipham} mipham)`) } catch { lines.push(t('commands.doctor.skills_unavailable')) } } // CLAUDE.md audit — flag sections the model can infer from the codebase try { const loader = new InstructionsLoader() loader.loadAll(process.cwd()) const claudeFiles = loader.list().filter((f) => f.path.endsWith('CLAUDE.md')) lines.push('') lines.push(t('commands.doctor.audit_section')) let found = false for (const file of claudeFiles) { const sections = findDerivableSections(file.content) if (sections.length === 0) continue found = true lines.push(` ${file.path}`) for (const s of sections) { lines.push(` - "${s.heading}" — ${DERIVABLE_HINTS[s.reason]}`) } } if (!found) { lines.push(t('commands.doctor.audit_clean')) } else { lines.push('') lines.push(t('commands.doctor.audit_found')) } } catch { // best-effort — skip the audit on any failure } return { content: lines.join('\n') } } const skillDoctorCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const lines: string[] = [t('commands.skillDoctor.title'), ''] const loader = ctx.skillsLoader if (!loader) { lines.push(t('commands.skillDoctor.unavailable')) return { content: lines.join('\n') } } const skills = loader.list() const usage = loadSkillUsage() // Per-skill context cost = the reminder entry estimate (matches buildSystemReminder). const rows = skills.map((s) => { const entry = `- ${s.name}: ${s.description}` return { name: s.name, type: s.type, tokens: Math.ceil(entry.length / 4) + 1, lastUsed: usage.get(s.name), } }) const unused = rows.filter((r) => r.lastUsed === undefined) const used = rows .filter((r) => r.lastUsed !== undefined) .sort((a, b) => b.lastUsed! - a.lastUsed!) const totalTokens = rows.reduce((sum, r) => sum + r.tokens, 0) const unusedTokens = unused.reduce((sum, r) => sum + r.tokens, 0) const age = (ts: number): string => { const days = Math.floor((Date.now() - ts) / 86400000) if (days <= 0) return t('commands.skillDoctor.today') return t('commands.skillDoctor.days_ago', { days: String(days) }) } lines.push( t('commands.skillDoctor.summary', { total: String(skills.length), tokens: String(totalTokens), unused: String(unused.length), unusedTokens: String(unusedTokens), }), ) lines.push('') if (unused.length > 0) { lines.push(t('commands.skillDoctor.unused_header')) for (const r of unused) { lines.push(` • ${r.name} (${r.type}) — ~${r.tokens} tokens`) } lines.push('') } if (used.length > 0) { lines.push(t('commands.skillDoctor.used_header')) for (const r of used) { lines.push(` • ${r.name} (${r.type}) — ${age(r.lastUsed!)}`) } } return { content: lines.join('\n') } } const fixCmd: CommandHandler = async (ctx, args) => { const t = resolveT(ctx) const { readFileSync, writeFileSync } = await import('node:fs') const { join } = await import('node:path') const { parse: parseYaml } = await import('yaml') const target = args.find((a) => a === 'doctor' || a === 'config' || a === 'cache' || a === 'test') const dryRun = args.includes('--dry-run') const apply = args.includes('--apply') const lines: string[] = [t('commands.fix.title')] if (dryRun) lines.push(t('commands.fix.dryrun_banner')) lines.push('') const readSafe = (p: string): string | null => { try { return readFileSync(p, 'utf-8') } catch { return null } } if (!target || target === 'doctor') { const loader = new InstructionsLoader() loader.loadAll(process.cwd()) const files = selectRepoClaudeFiles(loader.list()) const result = fixDoctor(files, { read: readSafe, write: (p, c) => { if (!dryRun) writeFileSync(p, c, 'utf-8') }, }) if (result.fixed.length === 0) { lines.push(t('commands.fix.doctor_clean')) } else { for (const f of result.fixed) { lines.push(t('commands.fix.doctor_fixed', { path: f.path, added: f.added.join(', ') })) } } lines.push('') } if (!target || target === 'config') { const configPaths = [join(process.cwd(), MIPHAM_DIR, 'config.yml'), miphamHome('config.yml')] const hookEngine = ctx.engine.getHookEngine?.() const result = fixConfig({ configPaths, read: readSafe, parseYaml, restore: (p) => tryRestoreFromBackup(p), hookHealth: () => hookEngine ? hookEngine.getHookHealth().map((h) => ({ key: h.key, disabled: h.health.disabled })) : [], reEnableHook: (k) => (hookEngine ? hookEngine.reEnableHook(k) : false), dryRun, }) if (result.corruptConfigs.length === 0 && result.disabledHooks.length === 0) { lines.push(t('commands.fix.config_clean')) } else { for (const p of result.restoredConfigs) { lines.push(t('commands.fix.config_restored', { path: p })) } for (const p of result.corruptConfigs.filter((p) => !result.restoredConfigs.includes(p))) { lines.push(t('commands.fix.config_corrupt', { path: p })) } for (const k of result.reenabledHooks) { lines.push(t('commands.fix.hook_reenabled', { key: k })) } for (const k of result.disabledHooks.filter((k) => !result.reenabledHooks.includes(k))) { lines.push(t('commands.fix.hook_disabled', { key: k })) } } lines.push('') } if (!target || target === 'cache') { const crsiDir = miphamHome('crsi') const cacheFiles = ['eval-scores.jsonl', 'improvements.jsonl', 'prose-proposals.jsonl'].map( (f) => join(crsiDir, f), ) const result = fixCache( cacheFiles, { read: readSafe, write: (p, c) => writeFileSync(p, c, 'utf-8'), }, apply && !dryRun, ) if (result.files.length === 0) { lines.push(t('commands.fix.cache_clean')) } else { for (const f of result.files) { lines.push( t('commands.fix.cache_found', { path: f.path, count: String(f.corruptLines.length) }), ) } if (apply && !dryRun) { for (const c of result.cleaned) { lines.push(t('commands.fix.cache_cleaned', { path: c.path, count: String(c.removed) })) } } else { lines.push(t('commands.fix.cache_hint')) } } lines.push('') } if (!target || target === 'test') { const testFile = args.find( (a) => !a.startsWith('--') && a !== 'test' && a !== 'doctor' && a !== 'config' && a !== 'cache', ) if (!testFile) { lines.push(t('commands.fix.test_usage')) } else { const llm = ctx.engine.getLlm() ?? ctx.engine.getRegistry() if (!llm) { lines.push(t('commands.fix.test_no_llm')) } else { const { execSync } = await import('node:child_process') const { dirname, resolve } = await import('node:path') const result = await fixCodeTarget( { runVitest: (file) => { try { const output = execSync(`pnpm vitest run '${file}'`, { cwd: process.cwd(), encoding: 'utf-8', stdio: ['ignore', 'pipe', 'pipe'], }) return { exitCode: 0, output } } catch (e) { const err = e as { stdout?: string; stderr?: string } return { exitCode: 1, output: `${err.stdout ?? ''}${err.stderr ?? ''}` } } }, readFile: readSafe, writeFile: (p, c) => writeFileSync(p, c, 'utf-8'), generateFix: async (prompt) => { let text = '' for await (const chunk of llm.chat({ model: '', messages: [{ role: 'user' as const, content: prompt }], temperature: 0, })) { if (chunk.type === 'text' && chunk.content) text += chunk.content } return stripCodeFences(text) }, resolveSourceFile: (tf, specifier) => { const base = resolve(dirname(tf), specifier) return /\.[cm]?[jt]sx?$/.test(base) ? base : `${base}.ts` }, }, testFile, { apply: apply && !dryRun, maxRetries: 3 }, ) if (result.fixed) { lines.push( t('commands.fix.test_fixed', { test: result.testFile, source: result.sourceFile ?? '-', attempts: String(result.attempts), }), ) if (!apply) lines.push(t('commands.fix.test_hint')) } else { lines.push( t('commands.fix.test_failed', { test: result.testFile, detail: result.detail ?? '' }), ) } } } lines.push('') } return { content: lines.join('\n').trimEnd() } } // ═══════════════════════════════════════════════════════════════ // GitHub & Git Workflow Commands (Claude Code parity) // ═══════════════════════════════════════════════════════════════ function gitDiffBridgeCmd(opts: { label: string noChangesKey: string runningKey: string errorKey: string forwardToAI: string | (() => string) }): CommandHandler { return async (ctx) => { const t = resolveT(ctx) try { const { execSync } = await import('node:child_process') const diff = execSync('git diff --stat', { encoding: 'utf-8', timeout: 5000 }).trim() if (!diff) { return { content: `─ ${opts.label} ─\n\n${t(opts.noChangesKey)}` } } return { content: `─ ${opts.label} ─\n\n${t(opts.runningKey)}\n\nChanged files:\n${diff}`, forwardToAI: typeof opts.forwardToAI === 'function' ? opts.forwardToAI() : opts.forwardToAI, } } catch { return { content: `─ ${opts.label} ─\n\n${t(opts.errorKey)}` } } } } const codeReviewCmd = gitDiffBridgeCmd({ label: 'Code Review', noChangesKey: 'commands.code_review.no_changes', runningKey: 'commands.code_review.running', errorKey: 'commands.code_review.error', forwardToAI: () => `use the code-review skill to review all uncommitted changes. Check all 7 dimensions: correctness, security, performance, code quality, architecture & design, testing, and language-specific issues. Use effort level: ${getPreference('lastCodeReviewEffort', 'high')}.`, }) const simplifyCmd = gitDiffBridgeCmd({ label: 'Simplify', noChangesKey: 'commands.simplify.no_changes', runningKey: 'commands.simplify.running', errorKey: 'commands.simplify.error', forwardToAI: 'use the self-review skill to review these uncommitted changes. Focus on 4 cleanup passes: 1) Reuse — find duplicated logic, replace with existing helpers; 2) Simplification — flatten nesting, remove redundant state and dead code; 3) Efficiency — fix repeated object creation, unnecessary I/O, memory issues; 4) Abstraction Level — ensure code sits at the right architectural layer. Apply equivalent transformations only — do NOT change logic or fix bugs.', }) const verifyCmd = gitDiffBridgeCmd({ label: 'Verify', noChangesKey: 'commands.verify.no_changes', runningKey: 'commands.verify.running', errorKey: 'commands.verify.error', forwardToAI: 'verify these uncommitted changes through runtime observation only. For each change: 1) Find the user-facing surface (CLI command, API endpoint, UI interaction); 2) Drive the changed code to execute; 3) Push boundaries — pass null, repeated values, wrong types, interrupt mid-flow (Ctrl-C), resize window; 4) Report verdict per change: PASS (works as expected), FAIL (does not work or breaks something), BLOCKED (cannot reach observable state), SKIP (no runtime surface, e.g. pure documentation). Do NOT run the test suite — observe real execution behavior only.', }) const designCmd: CommandHandler = (_ctx, args) => { const t = resolveT(_ctx) const topic = args.join(' ') || 'the current task' return { content: t('commands.design.start', { topic }), forwardToAI: `help me design the architecture for ${topic}. Explore 2-3 approaches with trade-offs, then present a design covering: component breakdown, data flow, interfaces between components, error handling strategy, and testing approach. Use the plan sub-agent if deeper analysis would help. Prefer simplicity — YAGNI.`, } } const lintCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) try { const { execSync } = await import('node:child_process') const { existsSync } = await import('node:fs') const { join } = await import('node:path') const cwd = process.cwd() const hasPackageJson = existsSync(join(cwd, 'package.json')) let lintOutput = '' if (hasPackageJson) { try { lintOutput = execSync( 'npx eslint . --ext .ts,.tsx,.js,.jsx --max-warnings 0 2>&1 || true', { encoding: 'utf-8', timeout: 30000, cwd, }, ).trim() } catch { lintOutput = '(ESLint not configured or not found)' } } else { lintOutput = '(No package.json found — not a Node.js project)' } const lines = lintOutput.split('\n').slice(0, 40) const truncated = lines.length >= 40 ? lines.join('\n') + '\n... (truncated)' : lines.join('\n') return { content: [ t('commands.lint.title'), '', truncated || '(no issues found)', '', hasPackageJson ? t('commands.lint.fix_hint') : t('commands.lint.setup_hint'), ].join('\n'), } } catch { return { content: t('commands.lint.error') } } } // ═══════════════════════════════════════════════════════════════ // Session Enhancement Commands (Claude Code parity) // ═══════════════════════════════════════════════════════════════ const filesCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const { readdirSync, statSync } = await import('node:fs') const { join } = await import('node:path') const cwd = process.cwd() try { const entries = readdirSync(cwd, { withFileTypes: true }) const items = entries .filter((e) => !e.name.startsWith('.') || e.name === MIPHAM_DIR || e.name === '.mcp.json') .slice(0, 40) .map((e) => { const icon = e.isDirectory() ? '📁' : '📄' try { const stat = statSync(join(cwd, e.name)) const size = stat.isFile() ? ` ${(stat.size / 1024).toFixed(1)}KB` : '' return ` ${icon} ${e.name}${size}` } catch { return ` ${icon} ${e.name}` } }) return { content: [ t('commands.files.title'), '', t('commands.files.cwd', { path: cwd }), '', ...items, entries.length > 40 ? t('commands.files.more_files', { count: String(entries.length - 40) }) : '', '', t('commands.files.hint'), ].join('\n'), } } catch { return { content: t('commands.files.error') } } } const statsCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const msgs = c.getMessages() const tokens = c.getEstimatedTokens() const tools = ctx.engine.getTools() const userMsgs = msgs.filter((m) => m.role === 'user').length const assistantMsgs = msgs.filter((m) => m.role === 'assistant').length const systemMsgs = msgs.filter((m) => m.role === 'system').length const base = [ t('commands.stats.title'), '', t('commands.stats.messages', { total: String(msgs.length), user: String(userMsgs), assistant: String(assistantMsgs), system: String(systemMsgs), }), t('commands.stats.tokens', { tokens: tokens.toLocaleString(), max: c.getMaxTokens().toLocaleString(), }), t('commands.stats.tools', { count: String(tools.size) }), t('commands.stats.provider', { provider: ctx.providerId }), t('commands.stats.model', { model: ctx.modelId }), t('commands.stats.permission', { permission: ctx.config.permission }), '', t('commands.stats.usage', { pct: ((tokens / c.getMaxTokens()) * 100).toFixed(1) }), ] // ── CRSI & SIS extensions ── const engine = ctx.engine const extras: string[] = [] try { const db = engine.getErrorSignatureDB?.() if (db) { const s = db.getStats() if (s.total > 0) { extras.push( '', '── SIS Self-Immune ──', `Signatures: ${s.total} total, ${s.active} active | Success rate: ${(s.avgSuccessRate * 100).toFixed(0)}%`, ) } } } catch { /* SIS not initialized */ } try { const autoMemory = engine.getAutoMemory?.() if (autoMemory) { const count = autoMemory.sessionReflectionCount ?? 0 if (count > 0) extras.push(`CRSI Reflections: ${count} turn(s) analyzed`) } } catch { /* CRSI not initialized */ } try { const meta = engine.getMetaRuleEngine?.() if (meta) { const analysis = meta.analyze() if (analysis.systemHealth) { extras.push( `System Health: ${analysis.systemHealth.score}/100 — ${analysis.systemHealth.assessment}`, ) } } } catch { /* Meta not initialized */ } return { content: [...base, ...extras].join('\n') } } const summaryCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) const c = ctx.engine.getContext() const msgs = c.getMessages() if (msgs.length === 0) { return { content: t('commands.summary.no_conversation') } } const userMsgs = msgs .filter((m) => m.role === 'user' && typeof m.content === 'string') .slice(0, 5) .map((m) => { const text = typeof m.content === 'string' ? m.content : '' return text.length > 100 ? text.slice(0, 100) + '...' : text }) return { content: [ t('commands.summary.title'), '', t('commands.summary.total_messages', { count: String(msgs.length) }), t('commands.summary.est_tokens', { tokens: c.getEstimatedTokens().toLocaleString() }), '', t('commands.summary.recent_topics'), ...userMsgs.map((t, i) => ` ${i + 1}. ${t}`), '', t('commands.summary.footer'), ].join('\n'), } } const cdCmd: CommandHandler = async (ctx, args) => { const t = resolveT(ctx) const target = args[0] if (!target) { return { content: t('commands.cd.usage', { cwd: process.cwd() }) } } const { existsSync } = await import('node:fs') const { resolve } = await import('node:path') const resolved = resolve(target.replace(/^~/, homedir())) if (!existsSync(resolved)) { const suggestions = suggestDirectories(resolved) let content = t('commands.cd.not_found', { path: resolved }) if (suggestions.length > 0) { content += `\n\n${t('commands.cd.suggestions')}\n${suggestions.map((s) => ` ${s}`).join('\n')}` } return { content } } try { process.chdir(resolved) try { const { SessionStore } = await import('../core/session-store') const saved = SessionStore.load(ctx.sessionId) if (saved) { const log = ctx.engine.getContext().getLog() if (log) { SessionStore.saveLog(ctx.sessionId, log, { provider: saved.metadata.provider, model: saved.metadata.model, cwd: resolved, }) } else { SessionStore.save(ctx.sessionId, saved.messages, { provider: saved.metadata.provider, model: saved.metadata.model, cwd: resolved, }) } } } catch { /* session persistence is best-effort */ } return { content: t('commands.cd.content', { path: resolved }) } } catch (err) { return { content: t('commands.cd.failed', { error: (err as Error).message }) } } } const hooksCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const { loadSettingsJson } = await import('../config/loader') const cwd = process.cwd() // This command *displays* the configured list, so it asks for the project file // explicitly — the loader's default drops project hooks (they are // repository-controlled code execution) and would hide hooks that really do // exist here. // // Two calls, one per source: the merged list cannot say which entry came from // which file, and the two are not governed alike — only the project file's // entries are gated on workspace trust. const userHooks = loadSettingsJson(cwd).hooks const projectHooks = loadSettingsJson(cwd, { includeProjectHooks: true }).projectHooks ?? {} // Listing them is not the same claim as their running. Project hooks are // gated, and this command renders them even when the gate is shut — so the // display has to name the source and say which entries will not run. const gated = !getWorkspaceTrust().isTrusted(cwd) type Entry = { matcher?: string; hooks: Array<{ type: string; command?: string }> } const sources: Array<{ project: boolean; hooks: Record }> = [ { project: true, hooks: projectHooks as Record }, { project: false, hooks: userHooks as Record }, ] // Event → its entries, project first — the order the merged list produced, so // a trusted workspace's output is unchanged by the split. const byEvent = new Map>() for (const { project, hooks } of sources) { for (const [eventName, entries] of Object.entries(hooks)) { if (!Array.isArray(entries) || entries.length === 0) continue const bucket = byEvent.get(eventName) ?? [] for (const entry of entries) bucket.push({ entry, project }) byEvent.set(eventName, bucket) } } if (byEvent.size === 0) { return { content: t('commands.hooks.no_hooks') } } const lines: string[] = [ t('commands.hooks.title'), '', t('commands.hooks.location', { path: '.mipham/settings.json' }), '', ] let count = 0 for (const [eventName, bucket] of byEvent) { for (const { entry, project } of bucket) { for (const hook of entry.hooks) { count++ const matcher = entry.matcher || '*' const cmd = hook.type === 'command' && hook.command ? hook.command : hook.type const source = t(project ? 'commands.hooks.source_project' : 'commands.hooks.source_user') const suffix = project && gated ? ` ${t('commands.hooks.gated')}` : '' lines.push(` [${source}] ${eventName} [${matcher}] → ${cmd}${suffix}`) } } } lines.push('') lines.push(t('commands.hooks.found', { count: String(count) })) return { content: lines.join('\n') } } const telemetryCmd: CommandHandler = async (ctx, args) => { const sub = (args[0] ?? 'status').toLowerCase() if (sub === 'on' || sub === 'off') { const enable = sub === 'on' setTelemetryEnabled(enable) if (enable) enableTelemetryNow() return { content: enable ? '✓ Telemetry enabled. Anonymous usage counts will be sent to the configured endpoint.' : '✓ Telemetry disabled. Nothing is collected and no queue file is written.', } } if (sub === 'reset-id') { const id = resetInstallId() return { content: `✓ New anonymous install id: ${id}` } } if (sub === 'endpoint') { const url = args[1] if (!url) return { content: 'Usage: /telemetry endpoint ' } setTelemetryEndpoint(url) return { content: url === NO_ENDPOINT ? '✓ Destination cleared. Telemetry stays on, but nothing is ever sent.' : `✓ Endpoint set to ${url}`, } } if (sub !== 'status') { return { content: 'Usage: /telemetry [status|on|off|reset-id|endpoint ]', } } // ── status ── const consent = getTelemetryConsent() ?? resolveTelemetry() const settings = readTelemetrySettings('user') // Two ways for `endpoint` to be empty, and they mean different things: never // resolved (the kill switch fired first) or deliberately cleared (the `none` // sentinel). A third — nothing configured — is unreachable now that the // default is a real URL, which is why this used to be a one-line fallback. const destination = consent.endpoint ? consent.endpoint : consent.endpointSource === 'off' ? '_(not resolved — telemetry is off)_' : '_(nothing — the `none` sentinel cleared it)_' const lines = [ '## 📡 Telemetry', '', `| Field | Value |`, `| --- | --- |`, `| State | ${consent.enabled ? '🟢 enabled' : '⚪ disabled'} |`, `| Decided by | \`${consent.source}\` |`, `| Endpoint | ${destination} |`, `| Source | \`${consent.endpointSource}\` |`, `| Install id | ${settings.installId ?? '_(not yet generated)_'} |`, `| Prompted | ${settings.promptedAt ?? '_(never)_'} |`, '', 'Full data dictionary: `docs/telemetry.md`', ] return { content: lines.join('\n') } } const hooksHealthCmd: CommandHandler = (ctx) => { const hookEngine = ctx.engine.getHookEngine?.() if (!hookEngine) { return { content: 'Hook engine not available.' } } const health = hookEngine.getHookHealth() if (health.length === 0) { return { content: '✅ All hooks are healthy — no failures recorded.' } } const lines: string[] = ['## 🪝 Hook Health Status', ''] for (const { key, health: h } of health) { const status = h.disabled ? '🔴 DISABLED' : h.failures > 0 ? '🟡 DEGRADED' : '🟢 HEALTHY' lines.push(`### ${status} \`${key}\``) lines.push(`| 连续失败 | ${h.failures} (threshold: 5) |`) lines.push(`| 总失败次数 | ${h.totalFailures} |`) if (h.disabled) { const remaining = Math.max(0, 5 - (Date.now() - h.disabledAt) / 60_000) lines.push(`| 自动恢复 | ${remaining.toFixed(0)} min remaining |`) lines.push('') lines.push(`Use \`/hooks enable "${key}"\` to re-enable manually.`) } lines.push('') } return { content: lines.join('\n') } } const hooksEnableCmd: CommandHandler = (ctx, args) => { const hookEngine = ctx.engine.getHookEngine?.() if (!hookEngine) return { content: 'Hook engine not available.' } const keyArg = args[0]?.trim().replace(/"/g, '') if (!keyArg) { return { content: 'Usage: /hooks enable ""\n\nUse /hooks health to see hook keys.' } } return hookEngine.reEnableHook(keyArg) ? { content: `✅ Hook "${keyArg}" re-enabled.` } : { content: `Hook "${keyArg}" not found or not disabled. Use /hooks health to check status.` } } const batchCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) return { content: t('commands.batch.title') } } // ═══════════════════════════════════════════════════════════════ // Export // ═══════════════════════════════════════════════════════════════ const exportCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const { writeFileSync } = await import('node:fs') const { join } = await import('node:path') const cwd = process.cwd() const timestamp = new Date().toISOString().replace(/[:.]/g, '-').slice(0, 19) const filename = `mipham-export-${timestamp}.md` const filepath = join(cwd, filename) const msgs = ctx.engine.getContext().getMessages() const lines: string[] = [ `# Mipham Code — Session Export`, `> ${new Date().toISOString()}`, `> Provider: ${ctx.providerId} / ${ctx.modelId}`, '', '---', '', ] for (const msg of msgs) { const roleLabel = msg.role === 'user' ? '🧑 User' : msg.role === 'assistant' ? '🤖 Mipham Code' : '⚠ System' const content = typeof msg.content === 'string' ? msg.content : JSON.stringify(msg.content) lines.push(`### ${roleLabel}`) lines.push('') lines.push(content) lines.push('') } writeFileSync(filepath, lines.join('\n'), 'utf-8') return { content: t('commands.export.confirmed', { path: filepath, count: String(msgs.length), lines: String(lines.length), }), copyContent: filepath, } } // ═══════════════════════════════════════════════════════════════ // Review — alias for /code-review (P1: 2026-08-06 polish) // ═══════════════════════════════════════════════════════════════ const reviewCmd = codeReviewCmd // ═══════════════════════════════════════════════════════════════ // PR Comments // ═══════════════════════════════════════════════════════════════ const prCommentsCmd: CommandHandler = async () => { try { const { execSync } = await import('node:child_process') // Get branch info const branch = execSync('git branch --show-current', { encoding: 'utf-8', timeout: 3000, }).trim() const mainBranch = execSync('git remote show origin 2>/dev/null | grep "HEAD branch" | cut -d: -f2', { encoding: 'utf-8', timeout: 3000, }).trim() || 'main' // Get diff stats vs main const diffStat = execSync( `git diff --stat origin/${mainBranch}...HEAD 2>/dev/null || git diff --stat ${mainBranch}...HEAD 2>/dev/null || echo "(no remote tracking)"`, { encoding: 'utf-8', timeout: 5000 }, ).trim() const commits = execSync( `git log --oneline origin/${mainBranch}..HEAD 2>/dev/null || git log --oneline ${mainBranch}..HEAD 2>/dev/null || echo "(no commits ahead)"`, { encoding: 'utf-8', timeout: 5000 }, ).trim() const lines: string[] = [ '─ PR Review ─', '', `Branch: ${branch}`, `Base: ${mainBranch}`, `Commits ahead: ${commits.split('\n').filter(Boolean).length}`, '', ] if (commits && commits !== '(no commits ahead)') { lines.push('Commits:') for (const c of commits.split('\n')) lines.push(` ${c}`) lines.push('') } if (diffStat && diffStat !== '(no remote tracking)') { lines.push('Changed files:') lines.push(diffStat) lines.push('') } lines.push( 'To generate PR description: type "write a PR description for these changes" in chat.', ) lines.push('To review PR: type "review this PR" or use /review.') return { content: lines.join('\n') } } catch { return { content: '─ PR Review ─\n\nCould not determine PR context. Are you in a git repository with a remote?', } } } // ═══════════════════════════════════════════════════════════════ // Resume // ═══════════════════════════════════════════════════════════════ const resumeLastCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const { SessionStore } = await import('../core/session-store') const latest = SessionStore.getLatest() if (!latest) { return { content: `${t('commands.resume.restored')}\n\n${t('commands.resume.no_sessions')}\n\n${t('commands.resume.empty_footer')}`, } } const session = SessionStore.load(latest.name) if (!session) { return { content: `${t('commands.resume.load_failed')}\n\nCould not load session "${latest.name}". The file may have been removed.`, } } const MAX_RESUME = 30 const truncated = session.messages.length > MAX_RESUME const messages = truncated ? session.messages.slice(-MAX_RESUME) : session.messages const date = new Date(latest.updatedAt).toLocaleString() const truncatedStr = truncated ? ` (showing last ${MAX_RESUME})` : '' return { content: [ t('commands.resume.restored'), '', t('commands.resume.restored_content', { name: latest.name, total: String(session.messages.length), truncated: truncatedStr, provider: latest.provider, model: latest.model, date, }), '', truncated ? t('commands.resume.restored_footer', { loaded: String(messages.length), total: String(session.messages.length), }) : t('commands.resume.restored_full_footer', { loaded: String(messages.length), // 文案里的 `{name}` 必须真给：`t()` 把没给的占位符替换成**空串** // （`t.ts` 的 `params[k] ?? ''`），漏掉就静默印出 `--resume ""`。 name: latest.name, }), ].join('\n'), forwardedMessages: messages, resumeWarning: true, } } const resumeDeleteCmd: CommandHandler = async (ctx, args) => { const t = resolveT(ctx) const name = args.join(' ').trim() if (!name) { return { content: t('commands.resume.delete_usage') } } const { SessionStore } = await import('../core/session-store') const deleted = SessionStore.delete(name) return deleted ? { content: t('commands.resume.delete_confirmed', { name }) } : { content: t('commands.resume.delete_not_found', { name }) } } const resumeCmd: CommandHandler = async (ctx, args) => { const t = resolveT(ctx) const sub = args[0]?.toLowerCase() if (sub === 'last') { return resumeLastCmd(ctx, args.slice(1)) } if (sub === 'delete') { return resumeDeleteCmd(ctx, args.slice(1)) } const { SessionStore } = await import('../core/session-store') const targetName = args.join(' ') if (targetName) { const session = SessionStore.load(targetName) if (session) { return { content: `${t('commands.resume.found_title')}\n\n${t('commands.resume.found_content', { name: session.metadata.name, messages: String(session.metadata.messageCount), provider: session.metadata.provider, model: session.metadata.model, updated: session.metadata.updatedAt, target: targetName })}`, } } return { content: `${t('commands.resume.not_found_title')}\n\n${t('commands.resume.not_found_content', { name: targetName })}`, } } const sessions = SessionStore.list() if (sessions.length === 0) { return { content: `${t('commands.resume.restored')}\n\n${t('commands.resume.no_sessions')}\n\n${t('commands.resume.empty_footer')}`, } } const recent = sessions.slice(0, 10) const lines: string[] = [ t('commands.resume.saved_title'), '', ...recent.map( (s, i) => ` ${(i + 1).toString().padStart(2)}. ${s.name.padEnd(45)} ${s.messageCount.toString().padStart(4)} msgs ${new Date(s.updatedAt).toLocaleString()}`, ), '', t('commands.resume.total_footer', { count: String(sessions.length) }), '', t('commands.resume.resume_hint'), t('commands.resume.resume_last_hint'), t('commands.resume.delete_hint'), t('commands.resume.cli_hint'), '', t('commands.resume.auto_save_hint'), ] return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Wiki Save // ═══════════════════════════════════════════════════════════════ const saveCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const target = args.join(' ').trim() const override = target ? `\nUser override: "${target}". If it names a note type (synthesis/concept/source/decision/session), file under that type; otherwise treat it as the note title.` : '' return { content: t('commands.save.content'), forwardToAI: `Invoke the save-to-wiki skill (use the Skill tool) to save this conversation into the Obsidian wiki.${override}`, } } // ═══════════════════════════════════════════════════════════════ // Memory Management // ═══════════════════════════════════════════════════════════════ const memoryCmd: CommandHandler = async (ctx, args) => { const t = resolveT(ctx) const { existsSync, readdirSync, readFileSync, statSync } = await import('node:fs') const { join } = await import('node:path') const memoryDir = miphamHome('memory') // /memory gc — 记忆卫生：归档「0 召回 + 过期」的 auto-* 记忆（手写只报告） if (args[0]?.toLowerCase() === 'gc') { const mm = getMemoryManager() const { archived, candidates } = mm.gc() const lines: string[] = [t('commands.memories.gc_title'), ''] if (archived.length === 0 && candidates.length === 0) { lines.push(t('commands.memories.gc_none')) } else { if (archived.length > 0) { lines.push(t('commands.memories.gc_archived', { count: String(archived.length) })) lines.push(...archived.map((n) => ` - ${n}`)) lines.push('') } if (candidates.length > 0) { lines.push(t('commands.memories.gc_candidates', { count: String(candidates.length) })) lines.push(...candidates.map((n) => ` - ${n}`)) lines.push('') } lines.push(t('commands.memories.gc_done')) } return { content: lines.join('\n') } } // /memory consolidate — 会话记忆合并：把 auto-* 聚簇成持久化 lesson-* if (args[0]?.toLowerCase() === 'consolidate') { const mm = getMemoryManager() const { merged, removed } = mm.consolidateAutoMemories() const lines: string[] = [t('commands.memories.consolidate_title'), ''] if (removed === 0) { lines.push(t('commands.memories.consolidate_none')) } else { lines.push( t('commands.memories.consolidate_result', { removed: String(removed), merged: String(merged), }), ) } return { content: lines.join('\n') } } // /memory dedup — 只读近重复报告：列出 TF-IDF 余弦 > 阈值的近重复对，用户决定是否合并 if (args[0]?.toLowerCase() === 'dedup') { const mm = getMemoryManager() const pairs = mm.listNearDuplicates() const lines: string[] = [t('commands.memories.dedup_title'), ''] if (pairs.length === 0) { lines.push(t('commands.memories.dedup_none')) } else { for (const { a, b, similarity } of pairs) { lines.push(` ${a} ↔ ${b} (${(similarity * 100).toFixed(0)}%)`) } lines.push('') lines.push(t('commands.memories.dedup_footer')) } return { content: lines.join('\n') } } if (!existsSync(memoryDir)) { return { content: t('commands.memories.no_memories') } } try { const files = readdirSync(memoryDir).filter((f) => f.endsWith('.md')) if (files.length === 0) { return { content: t('commands.memories.empty') } } const memories: Array<{ file: string; size: number; mtime: Date; title: string }> = [] for (const f of files) { const p = join(memoryDir, f) const stat = statSync(p) let title = f try { const content = readFileSync(p, 'utf-8') const match = content.match(/^#\s+(.+)$/m) if (match) title = match[1]! } catch { /* use filename */ } memories.push({ file: f, size: stat.size, mtime: stat.mtime, title }) } memories.sort((a, b) => b.mtime.getTime() - a.mtime.getTime()) const lines: string[] = [ t('commands.memories.title'), '', t('commands.memories.location', { path: memoryDir }), t('commands.memories.total', { count: String(memories.length), plural: memories.length === 1 ? 'y' : 'ies', }), '', ...memories.map( (m, i) => ` ${(i + 1).toString().padStart(2)}. ${m.file.padEnd(35)} ${(m.size / 1024).toFixed(1)}KB ${m.mtime.toLocaleDateString()} ${m.title}`, ), '', t('commands.memories.footer'), ] return { content: lines.join('\n') } } catch { return { content: t('commands.memories.error') } } } // ═══════════════════════════════════════════════════════════════ // Upgrade // ═══════════════════════════════════════════════════════════════ const upgradeCmd: CommandHandler = async (ctx) => { const t = resolveT(ctx) const { checkForUpdates, backupConfig, performUpdate, restoreConfig, getConfigPath } = await import('../shared/update') const update = checkForUpdates() // The registry was never reached. "Already up to date" would be a claim we // cannot support: `checkForUpdates` leaves `latest === current` on failure, so // the old code printed "✓ Already up to date (v0.81.8 → v0.81.8)" while // offline and gave no hint that a check had even been attempted. // (`mipham update` on the CLI path already says "Could not determine latest // version" — this is the TUI's equivalent.) if (!update.checked) { return { content: t('commands.upgrade.check_failed', { current: update.current }) } } if (!update.available) { return { content: t('commands.upgrade.uptodate', { current: update.current, latest: update.latest }), } } const backupPath = backupConfig(`upgrade-v${update.current}`) const lines: string[] = [ t('commands.upgrade.title'), '', t('commands.upgrade.available', { current: update.current, latest: update.latest }), '', ] if (backupPath) { lines.push(`Config backed up to: ${backupPath}`) } const result = performUpdate(update.latest) if (result.ok) { const configPath = getConfigPath() const { existsSync } = await import('node:fs') // 只有自证通过才置「已装待重启」—— 装坏了还提示重启，是在骗用户。 if (result.verified) ctx.setUpdateStatus({ state: 'installed', latest: update.latest }) lines.push('') lines.push(t('commands.upgrade.updated', { version: update.latest })) if (!result.verified) { lines.push(t('commands.upgrade.unverified', { reason: result.reason ?? '' })) } if (existsSync(configPath)) { lines.push(t('commands.upgrade.config_preserved', { path: configPath })) } else if (backupPath) { if (restoreConfig(backupPath)) { lines.push(t('commands.upgrade.config_restored')) } } lines.push('') lines.push(t('commands.upgrade.old_version_warning')) } else { lines.push('') // 失败信息必须回答两件事：为什么，以及**我手里还有没有 CLI**。 const rollbackNote = t( result.rolledBack ? 'commands.upgrade.rolled_back' : 'commands.upgrade.no_rollback', ) lines.push( t('commands.upgrade.update_failed', { reason: result.reason ?? '', command: NPM_UPDATE_COMMAND, path: backupPath || '', rolledBack: rollbackNote, }), ) } return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Language — i18n locale control // ═══════════════════════════════════════════════════════════════ const langCmd: CommandHandler = (ctx, args) => { const t = resolveT(ctx) const requested = args[0] if (!requested || !['en-US', 'zh-CN'].includes(requested)) { return { content: t('commands.lang.current', { locale: 'en-US' }) } } return { content: t('commands.lang.set', { locale: requested }) } } // ═══════════════════════════════════════════════════════════════ // No-Plan — exit plan mode // ═══════════════════════════════════════════════════════════════ const noPlanCmd: CommandHandler = (ctx) => { const t = resolveT(ctx) return { content: t('commands.no_plan.confirmed') } } // ═══════════════════════════════════════════════════════════════ // Workflows // ═══════════════════════════════════════════════════════════════ const workflowsCmd: CommandHandler = async () => { const { existsSync, readdirSync, readFileSync } = await import('node:fs') const { join } = await import('node:path') const locations = workflowScriptDirs(process.cwd()) const lines: string[] = ['─ Workflows ─', ''] let found = 0 for (const loc of locations) { if (!existsSync(loc)) continue try { const items = readdirSync(loc).filter((f) => f.endsWith('.js') || f.endsWith('.ts')) if (items.length === 0) continue lines.push(`📍 ${loc}`) for (const item of items) { const path = join(loc, item) try { const content = readFileSync(path, 'utf-8') const metaMatch = content.match( /export const meta\s*=\s*\{[^}]*name:\s*['"]([^'"]+)['"][^}]*description:\s*['"]([^'"]+)['"]/, ) if (metaMatch) { lines.push(` • ${item} — "${metaMatch[1]}" — ${metaMatch[2]}`) } else { lines.push(` • ${item}`) } } catch { lines.push(` • ${item}`) } found++ } } catch { /* skip */ } } if (found === 0) { lines.push('No workflow scripts found.') lines.push('') lines.push('Workflows are multi-agent orchestration scripts stored in:') lines.push(' .mipham/workflows/ (project-level — new scripts go here)') lines.push(' .claude/workflows/ (project-level, legacy — still read)') lines.push(' ~/.claude/workflows/ (user-level, legacy — still read)') lines.push('') lines.push('Create a .js file in either project location to add a workflow.') } else { lines.push('') lines.push(`${found} workflow(s) found.`) lines.push('Use /workflows to run a specific workflow.') } return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // /workflow — auto-generate + execute // ═══════════════════════════════════════════════════════════════ const workflowSaveCmd = async (name: string): Promise => { const { existsSync, mkdirSync, writeFileSync, readFileSync } = await import('node:fs') const { join } = await import('node:path') const targetDir = workflowScriptDir(process.cwd()) if (!existsSync(targetDir)) { mkdirSync(targetDir, { recursive: true }) } // Read the last-run state persisted by the Workflow tool. Every readable // location is checked, not just the writable one: a run that predates the // move to `.mipham/` left its state in `.claude/workflows/`, so looking in // the new directory alone would report "no recent run" on the first save // after upgrading. const stateFile = workflowScriptDirs(process.cwd()) .map((dir) => join(dir, '.last-run.json')) .find((file) => existsSync(file)) if (!stateFile) { return { content: 'No recent workflow run found. Run a workflow first with /workflow .' } } try { const state = JSON.parse(readFileSync(stateFile, 'utf-8')) const script = state.script as string if (!script) { return { content: 'No script found in last run state.' } } const safeName = name.replace(/[^a-zA-Z0-9_-]/g, '-') const scriptPath = join(targetDir, `${safeName}.js`) writeFileSync(scriptPath, script, 'utf-8') return { content: `Workflow saved to ${scriptPath}\nUse /workflow run ${safeName} to run it again.`, } } catch (err) { return { content: `Failed to save workflow: ${String(err)}` } } } const workflowRunCmd = async (name: string): Promise => { const { existsSync } = await import('node:fs') const { join } = await import('node:path') const safeName = name.replace(/[^a-zA-Z0-9_-]/g, '-') const locations = workflowScriptDirs(process.cwd()) for (const loc of locations) { const scriptPath = join(loc, `${safeName}.js`) if (existsSync(scriptPath)) { return { content: '', forwardToAI: `Read the workflow script at ${scriptPath}, then call the Workflow tool with ` + `the file contents as the "script" parameter to execute it. Report the results.`, } } } return { content: `Workflow "${safeName}" not found in .mipham/workflows/, ` + `.claude/workflows/ or ~/.claude/workflows/`, } } const workflowAutoCmd: CommandHandler = async (_ctx, args) => { const task = args.join(' ').trim() if (!task) { return { content: 'Usage: /workflow \n\n' + 'Describes the task, and the AI will generate a workflow script to execute it.\n' + 'Examples:\n' + ' /workflow audit all routes for missing auth\n' + ' /workflow research the impact of React 19 on our codebase\n' + ' /workflow find all hardcoded credentials in the codebase\n\n' + 'Sub-commands:\n' + ' /workflow save — save last successful workflow script\n' + ' /workflow run — run a saved workflow by name\n' + ' /workflows — list all saved workflow scripts', } } // Sub-commands const firstArg = args[0] || '' if (firstArg === 'save' && args[1]) { return workflowSaveCmd(args.slice(1).join(' ')) } if (firstArg === 'run' && args[1]) { return workflowRunCmd(args.slice(1).join(' ')) } // Default: forward to AI to generate + execute workflow return { content: '', forwardToAI: `Write and execute a workflow script for this task: ${task}\n\n` + `Use the Workflow tool to execute the generated script. ` + `After the workflow completes, summarize the results and offer to save the script ` + `with /workflow save if it is reusable.`, } } // ═══════════════════════════════════════════════════════════════ // /deep-research — multi-agent deep research pipeline // ═══════════════════════════════════════════════════════════════ const deepResearchCmd: CommandHandler = (_ctx, args) => { const topic = args.join(' ').trim() if (!topic) { return { content: stripIndent` Usage: /deep-research Conducts multi-agent deep research — 5-angle parallel search, adversarial verification, and top-tier synthesis into a well-cited report. Example: /deep-research impact of quantum computing on cryptography /deep-research latest advances in CRISPR gene editing `, } } return { content: stripIndent` ── Deep Research ── Topic: "${topic}" Orchestrating multi-agent research pipeline: 1. Scope — 5 research angles 2. Research — parallel web searches 3. Reduce — deduplicate + merge 4. Verify — adversarial skeptic pass 5. Synthesize — cited report, top model This may take a few minutes... `, forwardToAI: stripIndent` Conduct deep research on: "${topic}" Use the Workflow tool to orchestrate a multi-agent research pipeline: **Phase 1 — Scope**: Define 5 research angles covering: overview, technical-details, competitors/alternatives, criticism/limitations, future-trends. **Phase 2 — Research**: Fan out parallel web searches across all 5 angles. Each agent searches the web and returns sources with titles, URLs, and key points. Use schema validation to ensure structured output. **Phase 3 — Reduce**: Deduplicate sources by URL. Merge overlapping findings. EDGE LOGIC IS FREE — use plain JavaScript, not agent calls. **Phase 4 — Verify**: Run adversarial verification (verify with mode: 'adversarial', skeptics: 2, threshold: 1) on each key claim. Discard claims that fail verification. **Phase 5 — Synthesize**: Use the best available model to synthesize all verified findings into a comprehensive, well-cited research report. Include: - Executive summary - Key findings with source citations (title, URL) - Competing perspectives and disagreements - Gaps and unknowns - Recommendations for further reading Output the final report as a structured markdown document with inline citations. `, } } // ═══════════════════════════════════════════════════════════════ // Permissions // ═══════════════════════════════════════════════════════════════ // ═══════════════════════════════════════════════════════════════ // Recommend — analyze project and recommend setup (like Claude Code // automation recommender) // ═══════════════════════════════════════════════════════════════ // ═══════════════════════════════════════════════════════════════ // Setup — guided project initialization (mirrors Claude Code /setup) // ═══════════════════════════════════════════════════════════════ // ═══════════════════════════════════════════════════════════════ // Theme — display theme toggle // ═══════════════════════════════════════════════════════════════ const mcpCmd: CommandHandler = async (ctx, args) => { const client = McpClient.getInstance() const sub = args[0]?.toLowerCase() // /mcp connect if (sub === 'connect') { const name = args[1] if (!name) return { content: 'Usage: /mcp connect ' } const mcpServers = ctx.config.skills?.mcpServers ?? [] const config = mcpServers.find((s) => s.name === name) if (!config) { return { content: `Server "${name}" not found in config.\n\nConfigured: ${mcpServers.map((s) => s.name).join(', ') || '(none)'}`, } } if (config.auth?.type === 'oauth') { return { content: [ `── MCP Connect: ${name} (OAuth) ──`, '', 'Starting OAuth PKCE flow...', `Authorization: ${config.auth.authorizationUrl}`, `Scopes: ${config.auth.scopes?.join(', ') || '(default)'}`, ].join('\n'), forwardToAI: `Connect to MCP server "${name}" using OAuth. Call McpClient.getInstance().connectWithOAuth() with the server config, then register its tools. Report the result.`, } } const via = config.url ? 'HTTP' : 'stdio' if (config.url) { const headerKeys = config.headers ? Object.keys(config.headers) : [] const headerLine = headerKeys.length > 0 ? `\nHeaders to send: ${headerKeys.join(', ')}` : '' return { content: `── MCP Connect: ${name} ──\n\nConnecting via HTTP\nURL: ${config.url}${headerLine}\n\n⚠️ Verify the URL — any configured headers (e.g. Authorization) will be sent to this server.`, forwardToAI: `Connect to MCP server "${name}" using McpClient.getInstance().connect(config), then register its tools. Report the result.`, } } return { content: `── MCP Connect: ${name} ──\n\nConnecting via ${via}...`, forwardToAI: `Connect to MCP server "${name}" using McpClient.getInstance().connect(config), then register its tools. Report the result.`, } } // /mcp disconnect if (sub === 'disconnect') { const name = args[1] if (!name) return { content: 'Usage: /mcp disconnect ' } client.disconnect(name) // Dropping the connection is not dropping the tools. They live in the // engine's registry — the map the model actually calls — and closing the // transport leaves every one of them registered and selectable, failing // only at call time. Count what the registry actually gave up, not what the // connection used to hold: those are different objects and only one of them // is what "removed" refers to. const toolsMap = ctx.engine.getTools() const before = toolsMap.size unregisterMcpServerTools(name, toolsMap) const removed = before - toolsMap.size return { content: [ `── MCP Disconnect: ${name} ──`, '', removed > 0 ? `Disconnected. ${removed} tool(s) removed.` : 'Disconnected (no tools were registered).', ].join('\n'), } } // /mcp reconnect if (sub === 'reconnect') { const name = args[1] if (!name) return { content: 'Usage: /mcp reconnect ' } try { await client.reconnect(name) const count = client.getTools(name).length return { content: `── MCP Reconnect: ${name} ──\n\nReconnected — ${count} tool(s) rediscovered.`, } } catch (err) { return { content: [ `── MCP Reconnect: ${name} ──`, '', `Failed: ${String(err)}`, '', `The server stayed disconnected. Check that it is reachable, then retry — or use /mcp connect ${name}.`, ].join('\n'), } } } // /mcp reload if (sub === 'reload') { return { content: '── MCP Reload ──\n\nDisconnecting all and reconnecting...', forwardToAI: 'Disconnect all MCP servers via McpClient.getInstance().closeAll(), then reconnect all configured servers. Report each status.', } } // /mcp (default status) const configuredServers = ctx.config.skills?.mcpServers ?? [] const liveConnections = client.listConnections() const lines: string[] = ['── MCP Servers ──', ''] if (configuredServers.length > 0) { lines.push(`Configured servers (${configuredServers.length}):`) lines.push('') for (const s of configuredServers) { const live = liveConnections.find((c) => c.config.name === s.name) const statusIcon = live ? live.status === 'connected' ? '🟢' : live.status === 'connecting' ? '🟡' : live.status === 'error' ? '🔴' : '⚪' : '⚪' const statusLabel = live ? live.status : 'not started' const oauthTag = s.auth?.type === 'oauth' ? ' [OAuth]' : '' lines.push(` ${statusIcon} ${s.name}${oauthTag} [${statusLabel}]`) const transportInfo = s.url ? `URL: ${s.url}` : `Command: ${s.command ?? ''} ${(s.args ?? []).join(' ')}` lines.push(` ${transportInfo}`) if (live?.tools && live.tools.length > 0) { lines.push(` Tools: ${live.tools.length} registered`) } if (live?.error) lines.push(` Error: ${live.error}`) lines.push('') } } else { lines.push('No MCP servers configured.') } lines.push('── Commands ──') lines.push(' /mcp connect Connect to a server (OAuth or stdio)') lines.push(' /mcp disconnect Disconnect from a server') lines.push(' /mcp reconnect Reconnect a dropped server (with backoff)') lines.push(' /mcp reload Disconnect all and reconnect') lines.push('') lines.push('── Protocol ──') lines.push('MCP stdio transport (JSON-RPC 2.0) — fully implemented.') lines.push('Learn more: https://modelcontextprotocol.io') return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Phase 4 — Login / API Key Management // ═══════════════════════════════════════════════════════════════ const loginCmd: CommandHandler = (ctx) => { const activeProviders = ctx.config.providers.filter((p) => p.status === 'active') // Map provider IDs to their expected env var names const providerEnvMap: Record = { anthropic: 'ANTHROPIC_API_KEY', openai: 'OPENAI_API_KEY', deepseek: 'DEEPSEEK_API_KEY', google: 'GEMINI_API_KEY', qwen: 'QWEN_API_KEY', doubao: 'DOUBAO_API_KEY', hunyuan: 'HUNYUAN_API_KEY', } const lines: string[] = [ '── Authentication ──', '', 'Mipham Code uses API keys for authentication — no account login needed.', 'Each provider requires its own API key, set via environment variable or config file.', '', '── Provider API Keys ──', '', ] for (const p of activeProviders) { const envVar = providerEnvMap[p.id] ?? `${p.id.toUpperCase()}_API_KEY` const isSet = typeof process !== 'undefined' && !!process.env[envVar] const icon = isSet ? '✅' : '⬜' lines.push(` ${icon} ${p.id.padEnd(14)} $${envVar}${isSet ? ' (set)' : ''}`) } lines.push('') lines.push('── Setup ──') lines.push('') lines.push(' # Option 1: Environment variables (recommended)') lines.push(' export ANTHROPIC_API_KEY="sk-ant-..."') lines.push(' export OPENAI_API_KEY="sk-..."') lines.push('') lines.push(' # Option 2: Config file (~/.mipham/config.yml)') lines.push(' providers:') lines.push(' - id: anthropic') lines.push(' apiKey: $ANTHROPIC_API_KEY') lines.push('') lines.push('Current provider: ' + ctx.providerId + ' / ' + ctx.modelId) lines.push('') lines.push("Get API keys from each provider's developer console.") lines.push('Dashboard: https://mipham.ai/code/dashboard') return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Phase 4 — Logout / Clear Credentials // ═══════════════════════════════════════════════════════════════ const logoutCmd: CommandHandler = async () => { const { existsSync } = await import('node:fs') const userConfig = miphamHome('config.yml') const hasUserConfig = existsSync(userConfig) return { content: `── Sign Out ── Mipham Code uses API keys (not sessions), so there is no persistent login to "log out" of. To clear your credentials: 1. Unset environment variables (current session only): unset ANTHROPIC_API_KEY unset OPENAI_API_KEY unset DEEPSEEK_API_KEY # ... and others 2. Remove from shell profile (permanent): Edit ~/.zshrc or ~/.bashrc and remove the export lines. 3. User config: ${hasUserConfig ? '⚠ ~/.mipham/config.yml exists — check for stored keys' : '✅ ~/.mipham/config.yml not found (no stored keys)'} Note: Clearing keys will prevent Mipham Code from making API calls until you set them again with /login or manually. To switch providers without clearing keys, use /switch .`, } } // ═══════════════════════════════════════════════════════════════ // Phase 4 — Feedback // ═══════════════════════════════════════════════════════════════ const feedbackCmd: CommandHandler = async (ctx, args) => { const message = args.join(' ').trim() const lines: string[] = ['── Feedback ──', ''] // If a message was provided, save it to ~/.mipham/feedback/ if (message) { try { const { writeFileSync, mkdirSync, existsSync } = await import('node:fs') const { join } = await import('node:path') const dir = miphamHome('feedback') if (!existsSync(dir)) mkdirSync(dir, { recursive: true }) const ts = new Date().toISOString().replace(/[:.]/g, '-') const file = join(dir, `feedback-${ts}.md`) const content = [ '---', `timestamp: ${new Date().toISOString()}`, `version: ${ctx.version}`, `session: ${ctx.sessionId}`, `provider: ${ctx.providerId}`, `model: ${ctx.modelId}`, '---', '', message, ].join('\n') writeFileSync(file, content, 'utf-8') lines.push('Your feedback:') lines.push('') lines.push(' """') for (const line of message.split('\n')) { lines.push(' ' + line) } lines.push(' """') lines.push('') lines.push(`✅ Saved to \`${file}\``) lines.push('') } catch (err) { lines.push(`⚠️ Save failed: ${String(err)}`) lines.push('') } } lines.push('── Feedback Channels ──') lines.push('') lines.push(' 🐛 Bug Reports') lines.push(' GitHub Issues: https://github.com/One-Mipham/mipham-code/issues') lines.push(' Or run `/bug-report` to auto-generate a diagnostic report') lines.push('') lines.push(' 💡 Feature Requests') lines.push(' GitHub Issues: https://github.com/One-Mipham/mipham-code/issues') lines.push(' Template: Feature Request') lines.push('') lines.push(' 📧 General Feedback') lines.push(' Email: feedback@mipham.ai') lines.push(' Community: https://github.com/One-Mipham/mipham-code/discussions') lines.push('') lines.push('── System Info (include with bug reports) ──') lines.push(` Version: v${ctx.version}`) lines.push(` Provider: ${ctx.providerId} / ${ctx.modelId}`) lines.push(` Platform: ${process.platform} ${process.arch}`) lines.push( ` Runtime: ${typeof Bun !== 'undefined' ? 'Bun ' + Bun.version : 'Node.js ' + process.version}`, ) lines.push(` Node: ${process.version}`) return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Phase 4 — Agent Management // ═══════════════════════════════════════════════════════════════ const agentsCmd: CommandHandler = (ctx) => { const agentViewManager = ctx.engine.getAgentViewManager?.() if (!agentViewManager) { return { content: 'Agent View dashboard is initializing...' } } const counts = agentViewManager.countByStatus() const total = counts['needs-input'] + counts.working + counts.completed + counts.failed if (total === 0) { return { content: `── Agent View ── No background agents running. Use /bg to spawn a background agent session. Example: /bg Fix all TypeScript errors in the project /bg Review the auth module for security issues Or use the Agent tool in a conversation to launch a sub-agent.`, } } const lines: string[] = [ '── Agent View ──', '', ` ${total} session(s):`, ` 🟡 Needs input: ${counts['needs-input']}`, ` 🔵 Working: ${counts.working}`, ` 🟢 Completed: ${counts.completed}`, ` 🔴 Failed: ${counts.failed}`, '', 'Sessions:', ] const all = agentViewManager.list() for (const s of all.slice(0, 10)) { const statusIcon: Record = { 'needs-input': '🟡', working: '🔵', completed: '🟢', failed: '🔴', } const elapsed = s.elapsedMs < 1000 ? '<1s' : s.elapsedMs < 60000 ? `${Math.floor(s.elapsedMs / 1000)}s` : `${Math.floor(s.elapsedMs / 60000)}m` lines.push( ` ${statusIcon[s.status] || '⚪'} ${s.id} · ${s.provider}/${s.model} · ${elapsed} · ${s.task.slice(0, 50)}`, ) } if (total > 10) { lines.push(` ... and ${total - 10} more`) } lines.push('', 'Run `mipham agents` for the full interactive dashboard with j/k navigation.') return { content: lines.join('\n') } } const bgCmd: CommandHandler = async (ctx, args) => { const prompt = args.join(' ') if (!prompt.trim()) { return { content: `Usage: /bg Spawn a background agent to work on a task while you continue chatting. Examples: /bg Run the full test suite and report failures /bg Audit the src/ directory for security issues /bg Generate API documentation from JSDoc comments Background agents appear in the Agent View dashboard (/agents).`, } } const agentViewManager = ctx.engine.getAgentViewManager?.() if (!agentViewManager) { return { content: 'Agent View manager is initializing...' } } const session = agentViewManager.create(prompt, prompt, { provider: ctx.providerId, model: ctx.modelId, }) agentViewManager.addMessage(session.id, { role: 'user', content: prompt }) agentViewManager.updateStatus(session.id, 'working') session.kind = 'unattended' // A session row is not a task. Without the spawn below this command only ever // wrote a dashboard entry and reported success — the prompt was never handed // to a model, and the row sat at `working` forever. const bgReg = (await import('../agent/background-registry')).getBackgroundAgentRegistry() bgReg.spawn( prompt, 'general', async (_signal) => { const { SubAgent } = await import('../agent/sub-agent') const sa = new SubAgent( ctx.engine.getRegistry(), ctx.engine.getTools(), ctx.engine.getPermission(), undefined, undefined, ctx.engine.getLlm(), ) try { // The prompt came from the user typing `/bg …`, so it stays unframed — // unlike a workflow script's computed prompt. const result = await sa.execute(prompt, 'bg: ' + prompt.slice(0, 60), { type: 'general' }) agentViewManager.addMessage(session.id, { role: 'assistant', content: result || '(no output)', }) agentViewManager.updateStatus(session.id, 'completed') return result } catch (err) { // Leave the row honest on failure too — a session stuck at `working` // after its executor died is the same lie as never spawning at all. const message = err instanceof Error ? err.message : String(err) agentViewManager.addMessage(session.id, { role: 'assistant', content: `Background agent failed: ${message}`, }) agentViewManager.updateStatus(session.id, 'failed') throw err } }, 'unattended', ) return { content: `✓ Background agent spawned: ${session.id} Task: ${session.task.slice(0, 80)} Provider: ${session.provider} / ${session.model} Status: working Use /agents to view all background sessions. Run \`mipham agents\` for the interactive dashboard.`, } } // ═══════════════════════════════════════════════════════════════ // Artifacts // ═══════════════════════════════════════════════════════════════ const forkCmd: CommandHandler = async (ctx, args) => { const prompt = args.join(' ') if (!prompt.trim()) { return { content: [ 'Usage: /fork ', '', 'Fork a task into an isolated worktree with auto commit/push.', '', 'Workflow:', ' 1. Creates a new git worktree on branch worktree/', ' 2. Spawns a background agent to work on the task', ' 3. Auto-commits changes with Co-Authored-By on completion', ' 4. Auto-pushes to origin', ' 5. Reports branch and worktree path in the result', '', 'Examples:', ' /fork Refactor the auth module', ' /fork Add input validation to all API endpoints', ].join('\n'), } } const { execSync } = await import('node:child_process') const { join } = await import('node:path') try { execSync('git rev-parse --git-dir', { stdio: 'ignore' }) } catch { return { content: '/fork requires a git repository.' } } const agentViewManager = ctx.engine.getAgentViewManager?.() if (!agentViewManager) { return { content: 'Agent View manager is initializing...' } } const slug = prompt .toLowerCase() .replace(/[^a-z0-9]+/g, '-') .replace(/^-|-$/g, '') .slice(0, 40) const name = `${slug}-${Date.now().toString(36)}` const branch = `worktree/${name}` const wtPath = join(worktreeRoot(process.cwd()), name) try { execSync(`git worktree add -b ${branch} ${wtPath} HEAD`, { stdio: 'ignore', timeout: 30_000 }) } catch (err) { try { execSync(`git worktree remove --force ${wtPath}`, { stdio: 'ignore' }) } catch { /* ok */ } try { execSync(`git branch -D ${branch}`, { stdio: 'ignore' }) } catch { /* ok */ } return { content: `Worktree creation failed: ${err instanceof Error ? err.message : String(err)}`, } } const session = agentViewManager.create(prompt, prompt, { provider: ctx.providerId, model: ctx.modelId, }) agentViewManager.addMessage(session.id, { role: 'user', content: prompt }) agentViewManager.updateStatus(session.id, 'working') session.worktree = wtPath session.branch = branch session.kind = 'forked' const bgReg = (await import('../agent/background-registry')).getBackgroundAgentRegistry() bgReg.spawn( prompt, 'general', async (_signal) => { const { SubAgent } = await import('../agent/sub-agent') const sa = new SubAgent( ctx.engine.getRegistry(), ctx.engine.getTools(), ctx.engine.getPermission(), undefined, undefined, ctx.engine.getLlm(), ) const parentContext = ctx.engine.getContext() const result = await sa.execute(prompt, 'fork: ' + prompt.slice(0, 60), { worktreePath: wtPath, inheritContext: { messages: parentContext.getMessages() }, }) try { const { execSync: ex } = await import('node:child_process') ex(`git -C ${wtPath} add -A`, { stdio: 'ignore', timeout: 10_000 }) const st = ex(`git -C ${wtPath} status --porcelain`, { encoding: 'utf-8', timeout: 10_000 }) if (st.trim()) { ex(`git -C ${wtPath} commit -m "feat: ${prompt.slice(0, 60)}\n\n${COAUTHOR_TRAILER}"`, { stdio: 'ignore', timeout: 10_000, }) ex(`git push origin ${branch}`, { stdio: 'ignore', timeout: 30_000 }) } } catch { /* best-effort */ } return [ `## Fork done: ${prompt}`, '', `Branch: \`${branch}\``, `Worktree: \`${wtPath}\``, '', result || '(no output)', ].join('\n') }, 'forked', ) return { content: [ `✓ Forked: ${session.id}`, ` Task: ${prompt.slice(0, 80)}`, ` Branch: ${branch}`, ` Worktree: ${wtPath}`, ` Status: working`, '', 'Agent will auto-commit + push on completion.', 'Use /agents to track.', ].join('\n'), } } async function openBrowser(url: string): Promise { const { spawn } = await import('node:child_process') // Use spawn with array args — no shell, no command injection const [cmd, ...args] = process.platform === 'darwin' ? ['open', url] : process.platform === 'win32' ? ['cmd', '/c', 'start', '', url] : ['xdg-open', url] spawn(cmd!, args, { detached: true, stdio: 'ignore' }).unref() } const artifactOpenCmd: CommandHandler = async (ctx, args) => { const name = args[0] if (!name) { return { content: 'Usage: /artifact open \nExample: /artifact open dashboard' } } const { readManifest } = await import('../artifacts/manifest') const { artifactsRoot } = await import('../artifacts/paths') const manifest = readManifest(artifactsRoot(process.cwd())) // 先找本会话，再退到任意会话：文件在磁盘上跨会话留存，而 `open` 是用户手敲的。 // 原先写死 `session-1` + 端口 9876，工具回报的却是真会话 id 和**实际**端口 // （服务端遇到端口占用会自增），所以这条命令从构造上就打不开任何东西。 const entry = manifest.artifacts.find((a) => a.name === name && a.sessionId === ctx.sessionId) ?? manifest.artifacts.find((a) => a.name === name) if (!entry) { return { content: `✗ No artifact named "${name}". Run /artifact list to see them.` } } // 用工具落盘时记下的 URL —— 那就是它印给用户的同一个坐标，两边不该各算一次。 const url = entry.url try { await openBrowser(url) return { content: `✓ Opening artifact "${name}" in browser...\n ${url}` } } catch (err) { return { content: `✗ Failed to open browser: ${String(err)}\n URL: ${url}` } } } const artifactListCmd: CommandHandler = async (ctx, _args) => { const { readManifest, getSessionArtifacts } = await import('../artifacts/manifest') const { artifactsRoot } = await import('../artifacts/paths') const dir = artifactsRoot(process.cwd()) // 会话 id 取自真实上下文，不是写死的 'session-1' —— 工具落盘时写的是真 id， // 写死的那一支必然过滤出空列表。 const entries = getSessionArtifacts(dir, ctx.sessionId) if (entries.length === 0) { return { content: 'No artifacts created yet. Ask the AI to generate one with the Artifact tool.', } } const lines = ['── Artifacts ──', ''] for (const e of entries) { const ver = e.versions && e.versions.length > 1 ? ` v${e.versions.length}` : '' lines.push( ` ${e.name.padEnd(24)} ${(e.type + ver).padEnd(8)} ${(e.size / 1024).toFixed(1).padStart(8)}KB ${e.createdAt.slice(0, 16).replace('T', ' ')}`, ) } lines.push('', ` ${entries.length} artifact(s) — /artifact open to view`) // 画廊端口同样不写死：用最近一次落盘记下的那个（服务端会因端口占用自增）。 const { port } = readManifest(dir) if (port) lines.push(` Gallery: http://localhost:${port}`) return { content: lines.join('\n') } } const artifactServerCmd: CommandHandler = async (_ctx, _args) => { const url = 'http://localhost:9876' return { content: `── Artifact Server ──\n\n URL: ${url}\n Port: 9876\n\nArtifacts are served locally. No external network access.`, } } const artifactBaseCmd: CommandHandler = (ctx, args) => { const sub = args[0] if (sub === 'open') return artifactOpenCmd(ctx, args.slice(1)) if (sub === 'list') return artifactListCmd(ctx, []) if (sub === 'server') return artifactServerCmd(ctx, []) return { content: 'Usage: /artifact [name]\n\n /artifact open Open artifact in browser\n /artifact list List all artifacts\n /artifact server Show server status', } } // ═══════════════════════════════════════════════════════════════ // Remaining low-priority stubs (backend not yet available) // ═══════════════════════════════════════════════════════════════ // ═══════════════════════════════════════════════════════════════ // Commands — list all registered slash commands // ═══════════════════════════════════════════════════════════════ const commandsListCmd: CommandHandler = () => { const names = getCommandNames() const categories: Record = { 'Session & Identity': [], History: [], 'Model & Provider': [], 'Tools & Skills': [], Plugins: [], Workflow: [], Project: [], Environment: [], Account: [], Agents: [], Artifacts: [], Other: [], } const catMap: Record = { '/help': 'Session & Identity', '/version': 'Session & Identity', '/clear': 'Session & Identity', '/exit': 'Session & Identity', '/quit': 'Session & Identity', '/compact': 'Session & Identity', '/context': 'Session & Identity', '/status': 'Session & Identity', '/cost': 'Session & Identity', '/usage': 'Session & Identity', '/rename': 'Session & Identity', '/goal': 'Session & Identity', '/recap': 'Session & Identity', '/save': 'Session & Identity', '/export': 'Session & Identity', '/doctor': 'Session & Identity', '/skill-doctor': 'Session & Identity', '/dream': 'Session & Identity', '/constitution': 'Session & Identity', '/bug-report': 'Session & Identity', '/changelog': 'Session & Identity', '/resume': 'Session & Identity', '/resume last': 'Session & Identity', '/resume delete': 'Session & Identity', '/branch': 'Session & Identity', '/rewind': 'History', '/undo': 'History', '/copy': 'History', '/focus': 'History', '/ultracode': 'Workflow', '/pick': 'Model & Provider', '/model': 'Model & Provider', '/models': 'Model & Provider', '/ollama-refresh': 'Model & Provider', '/provider': 'Model & Provider', '/providers': 'Model & Provider', '/switch': 'Model & Provider', '/config': 'Model & Provider', '/fast': 'Model & Provider', '/effort': 'Model & Provider', '/theme': 'Model & Provider', '/upgrade': 'Model & Provider', '/tools': 'Tools & Skills', '/skills': 'Tools & Skills', '/reload-skills': 'Tools & Skills', '/browse-skills': 'Tools & Skills', '/install-skill': 'Tools & Skills', '/marketplace': 'Tools & Skills', '/browse-marketplace': 'Tools & Skills', '/remove-skill': 'Tools & Skills', '/mcp': 'Tools & Skills', '/plugins': 'Plugins', '/browse-plugins': 'Plugins', '/install-plugin': 'Plugins', '/remove-plugin': 'Plugins', '/plugin-enable': 'Plugins', '/plugin-disable': 'Plugins', '/commands': 'Tools & Skills', '/crsi rules': 'Tools & Skills', '/crsi disable': 'Tools & Skills', '/crsi analyze': 'Tools & Skills', '/crsi restore': 'Tools & Skills', '/crsi stats': 'Tools & Skills', '/crsi health': 'Tools & Skills', '/crsi inventory': 'Tools & Skills', '/crsi modify': 'Tools & Skills', '/crsi propose': 'Tools & Skills', '/crsi prose-clear': 'Tools & Skills', '/crsi eval': 'Tools & Skills', '/crsi bench': 'Tools & Skills', '/crsi meta': 'Tools & Skills', '/crsi interpret': 'Tools & Skills', '/crsi critique': 'Tools & Skills', '/crsi red-team': 'Tools & Skills', '/sis errors': 'Tools & Skills', '/sis stats': 'Tools & Skills', '/sis clear': 'Tools & Skills', '/plan': 'Workflow', '/no-plan': 'Workflow', '/tdd': 'Workflow', '/todos': 'Workflow', '/tasks': 'Workflow', '/review': 'Code Quality', '/pr-comments': 'Workflow', '/diff': 'Workflow', '/workflows': 'Workflow', '/workflow view': 'Workflow', '/workflow watch': 'Workflow', '/deep-research': 'Workflow', '/loop': 'Workflow', '/init': 'Project', '/setup': 'Project', '/permissions': 'Project', '/add-dir': 'Project', '/recommend': 'Project', '/security': 'Project', '/audit': 'Project', '/trust': 'Project', '/prompt-audit': 'Code Quality', '/ide': 'Environment', '/terminal-setup': 'Environment', '/memory': 'Environment', '/release-notes': 'Environment', '/login': 'Account', '/logout': 'Account', '/keys': 'Account', '/keys rotate': 'Account', '/keys audit': 'Account', '/keys view': 'Account', '/feedback': 'Account', '/telemetry': 'Account', '/agents': 'Agents', '/bg': 'Agents', '/fork': 'Agents', '/artifact': 'Artifacts', '/schedule': 'Other', '/commit': 'Workflow', '/push': 'Workflow', '/pr': 'Workflow', '/issue': 'Workflow', '/code-review': 'Code Quality', '/simplify': 'Code Quality', '/verify': 'Code Quality', '/design': 'Code Quality', '/lint': 'Code Quality', '/files': 'Session & Identity', '/stats': 'Session & Identity', '/summary': 'Session & Identity', '/cd': 'Session & Identity', '/hooks': 'Workflow', '/batch': 'Workflow', } for (const name of names) { const cat = catMap[name] ?? 'Other' categories[cat]?.push(name) } const lines: string[] = [`── All Slash Commands (${names.length}) ──`, ''] for (const [cat, cmds] of Object.entries(categories)) { if (cmds.length === 0) continue lines.push(`── ${cat} (${cmds.length}) ──`) for (const c of cmds) { lines.push(` ${c}`) } lines.push('') } lines.push('Type /help for details, or / to run.') return { content: lines.join('\n') } } // ═══════════════════════════════════════════════════════════════ // Command Registry // ═══════════════════════════════════════════════════════════════ const registry = new Map() // Session registry.set('/help', helpCmd) const pickCmd: CommandHandler = (ctx) => ({ content: `── Model Picker ── Use ↑↓ to navigate, Enter to select, Esc to close. Current: ${ctx.providerId} / ${ctx.modelId} Available: ${ctx.config.providers.filter((p) => p.status === 'active').length} providers, ${ctx.config.providers.flatMap((p) => p.models.filter((m) => m.status === 'active')).length} models Tip: Press Ctrl+P or type /pick from the chat input.`, }) registry.set('/pick', pickCmd) registry.set('/version', versionCmd) registry.set('/clear', clearCmd) registry.set('/exit', exitCmd) registry.set('/quit', exitCmd) registry.set('/compact', compactCmd) registry.set('/context', contextCmd) registry.set('/status', statusCmd) registry.set('/cost', costCmd) registry.set('/usage', usageCmd) registry.set('/rename', renameCmd) registry.set('/goal', goalCmd) registry.set('/recap', recapCmd) // History registry.set('/rewind', rewindCmd) registry.set('/undo', undoCmd) registry.set('/copy', copyCmd) registry.set('/focus', focusCmd) registry.set('/ultracode', ultracodeCmd) // Model & Provider registry.set('/model', modelCmd) registry.set('/models', modelsCmd) registry.set('/ollama-refresh', ollamaRefreshCmd) registry.set('/provider', providerCmd) registry.set('/providers', providersCmd) registry.set('/config', configCmd) registry.set('/fast', fastCmd) registry.set('/effort', effortCmd) registry.set('/switch', switchCmd) // Tools & Skills registry.set('/tools', toolsCmd) registry.set('/skills', skillsCmd) registry.set('/reload-skills', reloadSkillsCmd) registry.set('/browse-skills', browseSkillsCmd) registry.set('/install-skill', installSkillCmd) registry.set('/marketplace', marketplaceCmd) registry.set('/browse-marketplace', browseMarketplaceCmd) registry.set('/remove-skill', removeSkillCmd) registry.set('/plugins', pluginsCmd) registry.set('/browse-plugins', browsePluginsCmd) registry.set('/install-plugin', installPluginCmd) registry.set('/remove-plugin', removePluginCmd) registry.set('/plugin-enable', pluginEnableCmd) registry.set('/plugin-disable', pluginDisableCmd) registry.set('/commands', commandsListCmd) registry.set('/crsi rules', crsiRulesCmd) registry.set('/crsi disable', crsiDisableCmd) registry.set('/crsi analyze', crsiAnalyzeCmd) registry.set('/crsi restore', crsiRestoreCmd) registry.set('/crsi stats', crsiStatsCmd) registry.set('/crsi health', crsiHealthCmd) registry.set('/crsi inventory', crsiInventoryCmd) registry.set('/crsi modify', crsiModifyCmd) registry.set('/crsi propose', crsiProposeCmd) registry.set('/crsi prose-clear', crsiProseClearCmd) registry.set('/crsi eval', crsiEvalCmd) registry.set('/crsi bench', crsiBenchCmd) registry.set('/crsi meta', crsiMetaCmd) registry.set('/crsi interpret', crsiInterpretCmd) registry.set('/crsi critique', crsiCritiqueCmd) registry.set('/crsi red-team', crsiRedTeamCmd) registry.set('/sis errors', sisErrorsCmd) registry.set('/sis stats', sisStatsCmd) registry.set('/sis clear', sisClearCmd) registry.set('/sis cleanup', sisCleanupCmd) registry.set('/dream', dreamCmd) registry.set('/constitution', constitutionCmd) registry.set('/bug-report', bugReportCmd) registry.set('/changelog', changelogCmd) // Workflow registry.set('/plan', planCmd) registry.set('/tdd', tddCmd) registry.set('/todos', todosCmd) registry.set('/tasks', tasksCmd) registry.set('/diff', diffCmd) registry.set('/loop', loopCmd) registry.set('/no-plan', noPlanCmd) registry.set('/workflow', workflowAutoCmd) registry.set('/workflows', workflowsCmd) registry.set('/workflow view', workflowViewCmd) registry.set('/workflow watch', workflowWatchCmd) registry.set('/deep-research', deepResearchCmd) registry.set('/review', reviewCmd) registry.set('/pr-comments', prCommentsCmd) // Session Management registry.set('/doctor', doctorCmd) registry.set('/skill-doctor', skillDoctorCmd) registry.set('/fix', fixCmd) registry.set('/export', exportCmd) registry.set('/resume', resumeCmd) registry.set('/resume last', resumeLastCmd) registry.set('/resume delete', resumeDeleteCmd) registry.set('/memory', memoryCmd) registry.set('/save', saveCmd) registry.set('/upgrade', upgradeCmd) // Project registry.set('/init', initCmd) registry.set('/setup', setupCmd) registry.set('/recommend', recommendCmd) registry.set('/permissions', permissionsCmd) registry.set('/add-dir', addDirCmd) registry.set('/security', securityCmd) registry.set('/audit', securityCmd) registry.set('/trust', trustCmd) registry.set('/prompt-audit', promptAuditCmd) // Environment registry.set('/theme', themeCmd) registry.set('/lang', langCmd) registry.set('/ide', ideCmd) registry.set('/terminal-setup', terminalSetupCmd) registry.set('/release-notes', releaseNotesCmd) // Phase 4 — MCP, Auth, Feedback, Agents registry.set('/mcp', mcpCmd) registry.set('/login', loginCmd) registry.set('/logout', logoutCmd) registry.set('/keys', keysCmd) registry.set('/keys rotate', keysCmd) registry.set('/keys audit', keysCmd) registry.set('/keys view', keysCmd) registry.set('/feedback', feedbackCmd) registry.set('/agents', agentsCmd) registry.set('/bg', bgCmd) registry.set('/fork', forkCmd) // Artifacts registry.set('/artifact', artifactBaseCmd) // Lower-priority semi-stubs (WIP, backend pending) registry.set('/branch', branchCmd) registry.set('/schedule', scheduleCmd) // GitHub & Git Workflow (Claude Code parity) registry.set('/commit', commitCmd) registry.set('/push', pushCmd) registry.set('/pr', prCmd) registry.set('/issue', issueCmd) // Code Quality (Claude Code parity) registry.set('/code-review', codeReviewCmd) registry.set('/simplify', simplifyCmd) registry.set('/verify', verifyCmd) registry.set('/design', designCmd) registry.set('/lint', lintCmd) // Session Enhancement (Claude Code parity) registry.set('/files', filesCmd) registry.set('/stats', statsCmd) registry.set('/summary', summaryCmd) registry.set('/cd', cdCmd) registry.set('/hooks', hooksCmd) registry.set('/hooks health', hooksHealthCmd) registry.set('/hooks enable', hooksEnableCmd) registry.set('/telemetry', telemetryCmd) registry.set('/batch', batchCmd) // ═══════════════════════════════════════════════════════════════ // Public API // ═══════════════════════════════════════════════════════════════ export function getCommand(name: string): CommandHandler | undefined { return registry.get(name) } export function getCommandNames(): string[] { return Array.from(registry.keys()).sort() } /** * The commands `app.tsx` intercepts **before** the registry lookup. * * They return early and never reach `getCommand`, so the registry is not * authoritative for them. `/model-picker` is the one that is not a registry key * at all — the other five are, so listing them here is belt-and-braces rather * than a claim that they are missing. */ const PRE_REGISTRY_COMMANDS = ['/switch', '/pick', '/model-picker', '/exit', '/quit', '/focus'] /** * The bucket an unrecognised command name is recorded under. * * **Why this exists.** `parseSlashCommand` returns `parts[0]` — whatever the user * typed. Without a convergence point, `/foobar` mints a `command_calls./foobar` * series on the spot, and nothing bounds the number of series: * `MAX_LABEL_LENGTH` in `payload.ts` truncates the **value**, not the key count. * `tool_name` really is closed by construction (the registry declares the tool * set); `command_name` is not. */ export const UNKNOWN_COMMAND = '/unknown' /** The label a command name is recorded under: itself if we ship it, else the bucket. */ export function commandLabelFor(command: string): string { if (registry.has(command) || PRE_REGISTRY_COMMANDS.includes(command)) return command return UNKNOWN_COMMAND } /** * Every label `command_calls` can carry — the source the collector's allowlist is * generated from. * * Two more than `getCommandNames()`: `/model-picker` (user-typable, not a registry * key) and `UNKNOWN_COMMAND`. Miss either and the allowlist does not contain it, * so the collector folds those events into `__other__` — and `__other__` is * exactly what T4 must not be reading when it votes. */ export function getCommandLabelNames(): string[] { return Array.from(new Set([...registry.keys(), ...PRE_REGISTRY_COMMANDS, UNKNOWN_COMMAND])).sort() } export interface CommandEntry { name: string description: string } const COMMAND_DESCRIPTIONS: Record = { '/help': 'Show help', '/commands': 'List all slash commands', '/version': 'Show version info', '/clear': 'Clear conversation', '/exit': 'Exit Mipham Code', '/quit': 'Exit Mipham Code', '/compact': 'Compact context window', '/context': 'Show context stats', '/status': 'Session and system status', '/cost': 'Token usage estimate', '/usage': 'Detailed usage dashboard', '/rename': 'Rename current session', '/goal': 'Set session goal', '/recap': 'Summarize session so far', '/save': 'Save conversation to Obsidian wiki (skill: save-to-wiki)', '/export': 'Export conversation to file', '/doctor': 'System diagnostics', '/skill-doctor': 'Show unused skills and their context cost', '/fix': 'Deterministic self-repair: doctor/config/cache', '/dream': 'Background memory consolidation', '/constitution': 'View or reload constitutional principles', '/bug-report': 'Generate diagnostic report for GitHub Issues', '/changelog': 'Recent version history', '/resume': 'List saved sessions', '/resume last': 'Restore the most recent saved session', '/resume delete': 'Delete a saved session', '/branch': 'Fork conversation', '/rewind': 'Undo last AI turn', '/undo': 'Same as /rewind', '/copy': 'Copy last response to clipboard', '/focus': 'Toggle focus view', '/ultracode': 'Toggle multi-agent workflow orchestration mode', '/pick': 'Open model picker', '/model': 'Show current model', '/models': 'List all available models', '/provider': 'Show current provider', '/providers': 'List configured providers', '/switch': 'Switch provider and model', '/config': 'View configuration', '/fast': 'Toggle fast mode', '/effort': 'Set reasoning effort', '/theme': 'Set terminal theme', '/lang': 'Set display language (en-US, zh-CN)', '/tools': 'List available tools', '/skills': 'List loaded skills', '/reload-skills': 'Reload all skills', '/browse-skills': 'Browse community skill marketplace', '/install-skill': 'Install a skill by name or URL', '/marketplace': 'Manage marketplace sources (add/list/remove)', '/browse-marketplace': 'Browse skills across all marketplace sources', '/remove-skill': 'Remove an installed skill', '/mcp': 'MCP server status', '/plugins': 'List installed plugins', '/browse-plugins': 'Browse community plugin marketplace', '/install-plugin': 'Install a plugin from npm or local path', '/remove-plugin': 'Remove an installed plugin', '/plugin-enable': 'Enable a disabled plugin', '/plugin-disable': 'Disable an enabled plugin', '/crsi rules': 'List all active CRSI rules with their status', '/crsi disable': 'Disable a CRSI rule by ID', '/crsi analyze': 'Manually trigger CRSI pattern analysis across all agents', '/crsi restore': 'Restore a disabled or degraded CRSI rule', '/crsi stats': 'Show CRSI overall effectiveness statistics', '/crsi health': 'CRSI + SIS unified health dashboard with scoring', '/crsi inventory': 'Live capability self-report — CRSI/SIS/constitution state', '/crsi modify': 'Run a code self-modification through the sandbox (worktree → tests → approve)', '/crsi propose': '固化 CRSI 失败信号（默认教训 / --rule 受管理规则 / --prose 改 skill 散文），沙箱 + 人批准门控', '/crsi prose-clear': '清空散文提议去重 ledger（~/.mipham/crsi/prose-proposals.jsonl）', '/crsi eval': 'Run the ground-truth CRSI eval harness and record the score', '/crsi bench': 'Run the LLM task-performance benchmark and report the score', '/crsi meta': 'RSI Level 3 meta-rule analysis — rules that improve the rules', '/crsi interpret': 'Tool-call behavior dashboard — error patterns, usage, health', '/crsi critique': 'Enable/disable RLAIF self-critique on tool calls', '/crsi red-team': 'Run adversarial self-test — verify SIS blocks known attacks', '/sis errors': 'List all active SIS immune memory signatures', '/sis stats': 'Show SIS self-immune system aggregate statistics', '/sis clear': 'Retire an immune memory signature by ID', '/sis cleanup': 'Run immune memory garbage collection (dedup + cleanup)', '/plan': 'Enter plan mode', '/no-plan': 'Exit plan mode', '/tdd': 'Test-Driven Development workflow (RED → GREEN → REFACTOR)', '/todos': 'Task management (list/create tasks)', '/tasks': 'Background tasks', '/review': 'Review code for bugs, security, performance (alias for /code-review)', '/pr-comments': 'PR review summary', '/diff': 'Show git diff', '/workflows': 'List workflow scripts', '/workflow view': 'View workflow run details or list runs', '/workflow watch': 'Monitor active workflow execution', '/deep-research': 'Deep research with multi-agent parallel search, verification, and cited synthesis', '/loop': 'Run prompt on interval', '/init': 'Initialize .mipham config', '/setup': 'Guided project setup wizard', '/permissions': 'Show or persist permission rules', '/add-dir': 'Add workspace directory', '/recommend': 'Analyze project + recommend skills & setup', '/security': 'Security review checklist', '/audit': 'Same as /security', '/trust': 'Show and manage trusted workspaces', '/prompt-audit': 'Audit prompts for modern model optimization', '/ide': 'IDE integration guide', '/terminal-setup': 'Shell & terminal config', '/memory': 'Manage AI memories', '/release-notes': 'View version changelog', '/upgrade': 'Show upgrade instructions', '/login': 'Show API key status', '/logout': 'Clear credentials guide', '/keys': 'List API key rotation status', '/keys rotate': 'Rotate an API key', '/keys audit': 'Check for expired keys', '/keys view': 'View a provider plaintext API key', '/feedback': 'Send feedback', '/agents': 'Agent view dashboard', '/bg': 'Run a background agent task', '/fork': 'Fork task into isolated worktree', '/artifact': 'Manage artifacts', '/schedule': 'View scheduled tasks and cron jobs', '/commit': 'Generate commit message and review staged changes', '/push': 'Push the current branch', '/pr': 'Create a pull request', '/issue': 'File a GitHub issue', '/code-review': 'Review code for bugs, security, performance (7 dimensions)', '/simplify': 'Cleanup-only review: reuse, simplification, efficiency, abstraction', '/verify': 'Runtime verification — observe actual execution, not tests', '/design': 'Start architectural design session — explore approaches and trade-offs', '/lint': 'Run linting on the project', '/files': 'List files in current working directory', '/stats': 'Show session usage statistics', '/summary': 'Generate session summary', '/cd': 'Change session working directory', '/hooks': 'List configured hooks from settings.json', '/hooks health': 'Check hook health — see failures, disabled hooks, recovery status', '/hooks enable': 'Manually re-enable a hook that was auto-disabled after repeated failures', '/batch': 'Apply changes across multiple files', '/telemetry': 'Show or change anonymous usage-reporting settings', } export function getCommandList(): CommandEntry[] { return getCommandNames().map((name) => ({ name, description: COMMAND_DESCRIPTIONS[name] ?? '', })) } export function looksLikeSlashCommand(input: string): boolean { return input.trim().startsWith('/') } export function parseSlashCommand(input: string): { command: string; args: string[] } { const parts = input.trim().split(/\s+/) const command = parts[0]?.toLowerCase() || '' const args = parts.slice(1) return { command, args } } export { stripIndent } from './strip-indent.js' export { switchCmd as handleSwitch }