--- name: code-review description: One bounded, two-axis review of a fixed change set with typed findings and complete-coverage attestation. overrides: cwd,model,tier,thinking,concurrency,timeoutMs,maxTokens,maxGeneratedTokens result: code-review-v1 --- { "tasks": [ { "agent": "overwatch", "role": "standards", "task": "Perform exactly one read-only Standards review for this request:\n\n{task}\n\nFreeze and report the base and head commit identities before reviewing. Derive one changed-file manifest from that fixed range, inspect every path in it, and return one coverage record per path. Review only against repository-authored standards (AGENTS.md, CONTRIBUTING.md, documented architecture and test rules) and semantic correctness. Do not invoke /code-review, call flow, spawn agents, edit files, post to GitHub, or retry the review. A missing fixed point, unreadable path, or uncertain scope must be represented as skipped coverage or an unresolved question, never silently treated as clean.", "tools": "read,grep,find,ls,bash-ro", "contract": { "objective": "Complete one author-independent Standards review of the fixed change set described by: {task}", "constraints": [ "Read-only: do not change repository or external state.", "Freeze base/head identities and review every path in the resulting manifest.", "Return only evidence-backed semantic findings; do not manufacture style findings.", "Do not invoke another review workflow or agent." ], "nonGoals": [ "Fixing findings", "Posting GitHub comments", "Repeating until clean", "Refreshing unrelated CI evidence" ], "dependencies": [ "The caller-supplied task must identify a review target or fixed point." ], "authority": { "may": [ "Read repository files and documentation", "Run read-only git inspection commands", "Run bounded non-mutating verification commands when necessary" ], "mustNot": [ "Edit files", "Commit or push", "Post or resolve review comments", "Delegate or retry" ], "requiresApproval": [] }, "sideEffectClass": "read-only", "budget": { "timeoutMs": 1800000, "maxTokens": 300000, "maxGeneratedTokens": 12000 }, "acceptanceChecks": [ "axis is standards", "coverage contains every changed path with reviewed or skipped status", "every finding has a path, line anchor, claim, evidence, and suggested fix", "uncertainty appears as skipped coverage or unresolvedQuestions" ], "returnSchema": { "type": "object", "additionalProperties": false, "required": ["axis", "base", "head", "coverage", "findings"], "properties": { "axis": { "const": "standards" }, "base": { "type": "string", "pattern": "^[0-9a-f]{40,64}$" }, "head": { "type": "string", "pattern": "^[0-9a-f]{40,64}$" }, "coverage": { "type": "array", "items": { "type": "object", "additionalProperties": false, "required": ["path", "status", "evidence"], "properties": { "path": { "type": "string", "minLength": 1 }, "status": { "enum": ["reviewed", "skipped"] }, "evidence": { "type": "string", "minLength": 1 } } } }, "findings": { "type": "array", "items": { "type": "object", "additionalProperties": false, "required": ["id", "path", "startLine", "endLine", "severity", "category", "claim", "evidence", "suggestion"], "properties": { "id": { "type": "string", "minLength": 1 }, "path": { "type": "string", "minLength": 1 }, "startLine": { "type": "integer", "minimum": 1 }, "endLine": { "type": "integer", "minimum": 1 }, "severity": { "enum": ["critical", "high", "medium", "low"] }, "category": { "type": "string", "minLength": 1 }, "claim": { "type": "string", "minLength": 1 }, "evidence": { "type": "string", "minLength": 1 }, "suggestion": { "type": "string", "minLength": 1 } } } } } }, "owner": "standards" } }, { "agent": "overwatch", "role": "spec", "task": "Perform exactly one read-only Spec review for this request:\n\n{task}\n\nFreeze and report the base and head commit identities before reviewing. Derive one changed-file manifest from that fixed range, inspect every path in it, and return one coverage record per path. Verify the change against the issue, PR intent, acceptance criteria, and review-comment follow-ups named in the request. Do not invoke /code-review, call flow, spawn agents, edit files, post to GitHub, or retry the review. A missing specification, unreadable path, or uncertain scope must be represented as skipped coverage or an unresolved question, never silently treated as clean.", "tools": "read,grep,find,ls,bash-ro", "contract": { "objective": "Complete one author-independent Spec review of the fixed change set described by: {task}", "constraints": [ "Read-only: do not change repository or external state.", "Freeze base/head identities and review every path in the resulting manifest.", "Return only evidence-backed semantic findings tied to stated intent.", "Do not invoke another review workflow or agent." ], "nonGoals": [ "Fixing findings", "Posting GitHub comments", "Repeating until clean", "Refreshing unrelated CI evidence" ], "dependencies": [ "The caller-supplied task must identify the intended behavior and review target." ], "authority": { "may": [ "Read repository files and supplied issue or PR context", "Run read-only git inspection commands", "Run bounded non-mutating verification commands when necessary" ], "mustNot": [ "Edit files", "Commit or push", "Post or resolve review comments", "Delegate or retry" ], "requiresApproval": [] }, "sideEffectClass": "read-only", "budget": { "timeoutMs": 1800000, "maxTokens": 300000, "maxGeneratedTokens": 12000 }, "acceptanceChecks": [ "axis is spec", "coverage contains every changed path with reviewed or skipped status", "every finding has a path, line anchor, claim, evidence, and suggested fix", "missing or ambiguous intent appears in unresolvedQuestions" ], "returnSchema": { "type": "object", "additionalProperties": false, "required": ["axis", "base", "head", "coverage", "findings"], "properties": { "axis": { "const": "spec" }, "base": { "type": "string", "pattern": "^[0-9a-f]{40,64}$" }, "head": { "type": "string", "pattern": "^[0-9a-f]{40,64}$" }, "coverage": { "type": "array", "items": { "type": "object", "additionalProperties": false, "required": ["path", "status", "evidence"], "properties": { "path": { "type": "string", "minLength": 1 }, "status": { "enum": ["reviewed", "skipped"] }, "evidence": { "type": "string", "minLength": 1 } } } }, "findings": { "type": "array", "items": { "type": "object", "additionalProperties": false, "required": ["id", "path", "startLine", "endLine", "severity", "category", "claim", "evidence", "suggestion"], "properties": { "id": { "type": "string", "minLength": 1 }, "path": { "type": "string", "minLength": 1 }, "startLine": { "type": "integer", "minimum": 1 }, "endLine": { "type": "integer", "minimum": 1 }, "severity": { "enum": ["critical", "high", "medium", "low"] }, "category": { "type": "string", "minLength": 1 }, "claim": { "type": "string", "minLength": 1 }, "evidence": { "type": "string", "minLength": 1 }, "suggestion": { "type": "string", "minLength": 1 } } } } } }, "owner": "spec" } } ], "tier": "capable", "thinking": "high", "incompleteHandoffPolicy": "include", "concurrency": 2, "timeoutMs": 1800000, "maxTokens": 750000, "maxGeneratedTokens": 30000 }