# Changelog

All notable changes to this project are documented here. The format is based on
[Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to
[Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [0.3.3] — 2026-07-03

**An IDE on the floor, and a seventh engine.** Two headliners: a **built-in Monaco IDE** — the
VS Code editor engine in a full-window overlay, with a git CHANGES rail, side-by-side diffs vs
HEAD, a file tree, editor tabs, and Cmd/Ctrl+S save — and **GitHub Copilot CLI** as a first-class
agent engine, the project's **first community-contributed provider**
([PR #101](https://github.com/chaitanyagiri/munder-difflin/pull/101) by
[@anxkhn](https://github.com/anxkhn)).

### Added
- **Built-in Monaco IDE panel.** A title-bar **IDE** button toggles a full-window IDE overlay
  (matching the existing fullscreen-overlay pattern — the office floor, terminals, and voice UX are
  untouched). Left rail: a **git CHANGES list** (click a file → read-only **side-by-side diff vs
  HEAD**) plus the reused workspace **file tree** (click → edit). Right: **editor tabs** with
  dirty-state dots, save, and close; **Cmd/Ctrl+S** saves the active tab. The workspace root
  snapshots from the selected/god/first agent cwd. Monaco is **fully self-hosted** —
  electron-vite-safe bootstrap with bundled `?worker` imports and `loader.config({ monaco })`, no
  CDN — themed to the harness's light palette, and **all fs/git access goes through main-process
  IPC** (`git:diff` + preload bridge; the renderer holds no fs/git access)
  (`src/renderer/src/ide/*`, `src/main/git.ts`).
- **GitHub Copilot CLI agent engine** (`copilot`, npm `@github/copilot`) — community-contributed.
  Registered as a provider preset driven in Copilot's documented non-interactive **print mode**:
  `copilot -p "<prompt>" -s --allow-all-tools --no-ask-user [--model <id>]`, with the auto-approval
  flags **gated by the floor auto-mode toggle** like every other engine. Includes a
  **`COPILOT_MODELS` picker** (Claude Sonnet 4.5 default · GPT-5.4 · auto), `--resume` session
  continuity (best-effort), voice-hire (`spawn`) support, binary inference for pasted commands, and
  the official `npm install -g @github/copilot` offered by the missing-CLI installer. Non-hiveAware
  by design: print mode exits per turn and exposes no hook bridge, so `canReceiveInbox` is `false`
  and routed mail bounces to the GOD orchestrator instead of silently dropping
  (`src/shared/agentProvider.ts`, PR #101 — thanks [@anxkhn](https://github.com/anxkhn)).
- **Agent-provider registry test.** A self-contained, framework-free test
  (`node test/agent-provider.test.cjs`) transpiles the shared registry and asserts provider
  selection, the copilot command shape, model/resume passthrough, and codex non-regression.

### Fixed
- **IDE: no silent loss of edits typed during an in-flight save.** Keystrokes entered while a save
  was still writing can no longer be dropped when the save completes.

## [0.3.2] — 2026-06-27

**Talk to Michael.** The headline is **Realtime Michael** — a low-latency **voice channel to the
GOD orchestrator**, running alongside the async terminal floor. Press **Talk**, and Michael listens,
answers, and *acts* in real time: he reads the hive (tasks, board, memory, agents, activity) and —
behind spoken **echo-back confirmation** for anything destructive — creates and assigns work,
dispatches agents, spawns and kills workers, and steers the floor, all attributed to a distinct
**michael-voice** actor that pings the GOD terminal. He greets you on connect, **speaks task
completions the moment they land** ("respond when done"), and runs under a live cost meter with a
hard spend cap and an idle auto-disconnect. It's **bring-your-own OpenAI key**: the key is decrypted
**main-only**, minted into short-lived ephemeral session tokens, and never reaches the renderer. Plus
**Slack hardening** (proactive posting off by default; no sends without an explicit thread), a
dedicated **auto-compact** maintenance schedule decoupled from missions, and **per-agent environment
metadata**.

> **Live verification note.** The realtime voice loop is **human-verified end-to-end** on a real
> OpenAI key — connect → mic → Michael answers via the read tools, and the full destructive action
> path (spoken echo-back confirm → spawn / kill / dispatch → the worker appears on the floor →
> completion spoken back) was exercised live. It requires **your own OpenAI key with Realtime API
> access**; without one the **Talk** button stays visibly disabled with a "needs OpenAI key" cue.

### Added
- **Realtime Michael — talk to the GOD orchestrator by voice.** A new low-latency realtime channel
  (OpenAI Realtime API over WebRTC) sits next to the async terminal. A **Talk** toggle (on Michael's
  card and in any fullscreen terminal) opens a mic session with EC/NS/AGC, semantic-VAD turn-taking +
  barge-in, and a device picker for both microphone and speaker. Michael runs his own persona and
  answers in a natural voice, with an `Off → Connecting → Listening → Responding → Working` state
  machine surfaced live on his card (`src/renderer/src/realtime/*`, `src/main/realtime.ts`,
  `RealtimeMichaelToggle.tsx`).
- **BYOK ephemeral-token mint.** The voice session authenticates with a **short-lived ephemeral
  client secret** minted main-side from your stored OpenAI key — the real key is decrypted main-only
  and never crosses IPC. The renderer's CSP allows the WebRTC SDP exchange to reach `api.openai.com`
  while keeping everything else locked down (`src/main/realtime.ts`, renderer `index.html` CSP).
- **Voice action set with tiered echo-back confirmation.** Michael can *do* things by voice — read
  tools (tasks / board / memory / agents / activity / cost) plus the full action set: create and
  assign tasks, dispatch agents, pause / steer / halt, spawn / hire, kill, and edit schedules. Every
  **destructive** verb is gated behind a spoken **echo-back confirmation** (a distinct confirm token,
  never a bare "yes"), with hard refusals for killing the GOD agent or targeting all agents at once
  (`src/renderer/src/realtime/actions.ts`, `src/main/realtimeActions.ts`).
- **"Respond when done" completion loop.** Voice-dispatched work reports back on its own: a
  main-process completion watcher detects when a dispatched task finishes (card → done or a done
  reply in the inbox) and **proactively pushes the event into the live session so Michael speaks it
  unprompted**, while a `CompletionToast` shows it on screen. If the session is closed, completions
  queue to a desktop notification and a "completions since last session" warm-start; a `wait_for`
  tool covers the block-until-done case (`src/main/realtimeCompletionWatcher.ts`,
  `CompletionToast.tsx`).
- **michael-voice as a distinct actor.** Actions taken by voice are attributed to a separate
  **michael-voice** identity in messages, the board, and the activity log, and notify the GOD PTY —
  so a voice-driven dispatch is auditable and never silently impersonates a worker.
- **Cost guard + idle auto-disconnect for voice sessions.** A live session cost HUD by the Talk
  toggle, a configurable **spend cap** that auto-disconnects when hit, and a configurable **idle
  auto-disconnect** (default 3 min, 30 s–10 min or Off) so a forgotten-open mic can't run up a bill
  (`src/renderer/src/realtime/cost*`, **Settings → AI Engines**).
- **Greeting on connect.** When a session goes live, Michael opens with a warm, rotating greeting
  ("Hi, what's up?", "Hey, how's it going?", …) instead of waiting in silence — best-effort and
  guarded so a not-yet-ready data channel never blocks a successful connect.
- **Conversational read-layer.** The voice read tools were reworked to actually answer hive
  questions: `get_memory` no longer dead-ends, and new agent/board tools plus an expanded persona let
  Michael talk through roster, tasks, and floor state naturally.
- **Voice read-layer over hive messages (read/brief-only).** Realtime Michael can now read message
  *content*, not just metadata: a `get_messages` tool returns a **full message by id, one mailbox, or
  the latest across the floor** to brief the operator. **All redaction is main-side** —
  `voiceMessages()` runs every `subject`/`body` through `redactSecrets()` before the result leaves
  the main process, so the renderer/voice layer only ever receives already-redacted bodies (no
  provider / Slack / GitHub / AWS / Google key, JWT, PEM private-key block, or `Bearer` token can
  leak) and holds **zero** redaction policy. Read-only: it adds no write/mutate path — voice writes
  still go through the separate confirm-gated action spine (`src/main/hive.ts` `voiceMessages` +
  `redactSecrets`, `src/renderer/src/realtime/{tools.ts,VOICE-MESSAGE-ACCESS.md}`,
  `test/voice-messages.test.cjs`).
- **Talk reachable from any fullscreen terminal.** The Talk toggle is no longer Michael-only chrome —
  it's reachable in any fullscreen terminal view (the toggle is global/session state, so it's correct
  everywhere), while the per-session cost HUD stays Michael-only (`FullscreenTerminal.tsx`).
- **OpenAI Realtime key — documented and gated.** **Settings → AI Engines** now documents the
  **OpenAI Realtime key** as its own requirement (the same OpenAI provider key, distinct from your
  Anthropic key; main mints a short-lived token from it per session). The Talk button shows a live
  enabled/disabled status and an inline **"needs OpenAI key"** cue when none is set, so connecting
  never lands on a silently-dead button (`RealtimeMichaelToggle.tsx`, `SettingsModal.tsx`,
  `AiEnginesSettings.tsx`).
- **Dedicated auto-compact maintenance schedule.** Auto-compaction is now a **persistent,
  configurable maintenance mission** of its own, decoupled from the standup mission it used to ride
  on (so editing standups can't silently drop it). It reappears disabled rather than vanishing, with
  a mandatory warning and a configurable interval in the **Schedules** tab, plus a migration
  (`src/main/schedules*`, `SchedulesTab`).
- **Per-agent environment metadata + cwd guard.** Each agent now carries queryable environment
  metadata with a working-directory validity guard, and a new agent-env query tool (`src/main/hive.ts`).

### Changed
- **"Voice" is now "Talk".** The voice feature is renamed **Talk** throughout, with a redesigned
  navigation: the GOD card pops with a dedicated **Talk** line, and the worker nav cards are
  compacted to make room (`src/renderer/src/components/*`).
- **Robust voice task-matching (findCard).** Resolving a task by voice is now tolerant of
  hyphens/punctuation, phrasing, and truncation: both the spoken phrase and the stored title are
  normalized, candidates are **scored** (exact / prefix / token-coverage / substring), and close
  matches trigger a spoken **"which one?"** disambiguation instead of silently mutating the wrong card
  (`src/main/realtimeActions.ts`, `test/realtime-findcard.test.cjs`).

### Fixed
- **Fullscreen agent modal now opens above the fullscreen view.** The Add-Agent modal launched from
  the in-fullscreen "+ agent" button rendered *behind* the fullscreen terminal (z-index 100 vs 250)
  and was non-interactive. It's lifted to the dialog tier (300) so it's on top and clickable, and the
  fullscreen Esc handler now closes the modal first instead of exiting fullscreen underneath it
  (`AddAgentModal.tsx`, `FullscreenTerminal.tsx`).
- **Voice `get_memory` no longer dead-ends.** The conversational read tools were fixed so memory
  lookups return usable answers instead of stalling the turn.

### Security
- **BYOK voice secret invariant.** The real OpenAI key is encrypted at rest and decrypted
  **main-only**; the renderer only ever sees a **short-lived ephemeral client secret** minted per
  session. The key never crosses IPC and is never logged. The voice read-layer reports **tokens, not
  dollars** (de-monetized chrome), and every destructive voice action is held behind spoken echo-back
  confirmation with hard refusals for killing the GOD agent or targeting all agents at once.
- **Slack: proactive posting off by default + explicit-thread guard.** App/voice Slack sends are now
  **off by default** behind a config flag + Settings toggle, and a request with no explicit
  channel+thread is **refused** rather than guessed — closing an unattended-broadcast path
  (`src/main/slack.ts`, Settings).

## [0.3.1] — 2026-06-22

Three more coding CLIs join the floor — **OpenCode**, **Crush**, and **pi.dev** — each usable as a
worker *and* as Michael, with **bring-your-own keys + local LLMs**. Plus two reliability fixes: the
sleep-frozen message router and Codex workers' filesystem permissions.

> **Live verification note.** The three engines are wired end-to-end and selectable as god, and
> their architecture (preset + bridge + payload contract) was reviewed line-by-line. Their bridges'
> *runtime* behavior needs real model calls, so the following are **on-device checks pending BYOK
> keys / a local LLM** (not runtime-proven here):
> 1. each bridge's **turn-end signal** actually fires — OpenCode `session.idle`, pi `agent_end`,
>    Crush's proxy-synthesized `Stop` — flipping the agent to *idle*;
> 2. **OpenCode local-LLM** happy path: pick `local/<id>` with a base-URL set and confirm a turn
>    completes (the injected config now registers the *selected* model id);
> 3. **Crush** routes through the proxy on an OpenAI-wire model (the default god is now
>    `openai/gpt-4o`) and Crush honors the partial `base_url` override;
> 4. the **auto-mode gate** holds (no `permission:allow` / `--yolo` when the floor toggle is off).
>
> Crucially, mail delivery does **not** depend on those signals: a new **provider-agnostic
> PTY-quiescence idle fallback** flips any silent-but-pinned-`working` agent to idle, so the
> provider-agnostic idle inbox-wake nudge drains a god even if a bridge's turn-end signal never
> fires. That backstop is the safety net under shipping all three as `canReceiveInbox:true`.

### Added
- **Three new selectable engines: OpenCode · Crush · pi.dev.** Each lands as a declarative
  `AgentProviderPreset` and appears automatically in the Add-Agent picker (worker) and the god
  engine picker (orchestrator). Each gets a **bridge** for live status + turn-end inbox-drain:

  | Engine | Identity | Bridge | Notes |
  |---|---|---|---|
  | **OpenCode** | `opencode` (anomalyco/opencode, TS) | **native plugin** (`session.idle`) | bundled per-agent plugin; no traffic interception; auto-approve via gated `OPENCODE_CONFIG_CONTENT` |
  | **Crush** | `crush` (charmbracelet, Go) | **proxy** (qwen-tier) | per-agent `CRUSH_GLOBAL_CONFIG` routes traffic through the loopback sidecar (Crush has no base-URL env) |
  | **pi.dev** | `pi` (earendil-works) | **hooks** (bundled extension) | `pi.on(event)` → HIVE_SOCK; extension auto-approves tools only when the floor is in auto mode |

- **BYOK + local-LLM config UI (Settings → AI Engines).** A new per-provider config surface: API
  keys for the backend model-providers (Anthropic / OpenAI / Google / OpenRouter / Groq) stored
  **write-only** in the encrypted secret broker (never read back to the renderer; materialized
  main-only at spawn), plus per-engine **local base-URL** + default-model fields
  (`HarnessConfig.providerBaseUrls` / `providerDefaultModels`). Pi/OpenCode/Crush/Qwen pick up the
  keys + endpoints at spawn; auto-mode stays gated behind the floor toggle, and each engine runs
  unsandboxed in auto mode (surfaced as a caveat).
- **Provider-agnostic idle backstop (PTY-quiescence fallback).** A floor-wide check flips any agent
  pinned `working` with no terminal output for a short window back to *idle* — so the idle
  inbox-wake nudge can always drain a non-Claude god even if its bridge's turn-end signal (Stop /
  `session.idle` / `agent_end`) never fires. This is the safety net under shipping all three engines
  as god-eligible (`canReceiveInbox:true`) while their bridges await on-device verification
  (`src/renderer/src/hooks/useHive.ts`).

### Fixed
- **Codex hive workers get full filesystem + auto-approval from spawn (parity with Claude).** A Codex-engine agent in auto mode launched with `-a never -s workspace-write`, whose sandbox scopes writes to the PTY cwd (the user's project). But a hive worker must also write to its agent folder at `<harnessHome>/hive/agents/<id>/` (move `inbox/` → `.done/`, append `memory.md`, drop outbox JSON, write deliverables) — a **different path tree from cwd**, which `workspace-write` blocked. So a freshly spawned Codex worker couldn't complete HIVE PROTOCOL housekeeping and reported "it does not have permissions … grant write permission to the agent folder." Codex's auto-mode flag is now `--dangerously-bypass-approvals-and-sandbox` — the documented equivalent of Claude's `bypassPermissions` / Antigravity's `--dangerously-skip-permissions` (skip all approval prompts **and** drop the OS sandbox), so a Codex worker has the same filesystem access and auto-approval as a Claude worker from the get-go (`src/shared/agentProvider.ts`; reference/copy updated in `src/shared/codexCommands.ts`, `OnboardingWizard.tsx`, `renderer/store/config.ts`). Claude/agy/antigravity behavior is unchanged.
- **Re-arm the hive message router on wake (god→worker delivery survives sleep).** The outbox→inbox router is a `setInterval` (`hive.routeOnce` every ~1.5s) that, like the always-on beats, freezes during true macOS system sleep. `onSystemResume()` already re-armed the mission scheduler, the fleet/breaker beats, and keep-awake on `powerMonitor` `resume`/`unlock-screen` — but it never re-armed the router. So after a long sleep (e.g. laptop closed overnight) the scheduler→god path recovered while **every agent's outbox silently stopped draining**: god→worker, worker↔worker, and broadcast mail piled up undelivered, and no `message` event was logged. The resume handler now re-arms the router (clear-then-set, idempotent) **and** immediately drains the accumulated backlog instead of waiting for the first post-wake tick; the renderer's idle inbox-wake nudge then wakes each parked recipient once its mail lands (`src/main/index.ts`). Verified by `scripts/verify-keepalive-catchup.mjs` (now also reproduces the pre-fix backlog stall and proves the re-arm + flush).
- **Open-source model quick-picks + local-setup guides in Add-Agent.** Hiring a worker on a local-capable CLI engine (OpenCode/Crush/pi.dev) now shows curated **OSS-model quick-picks** — a **Local** bucket (Mac-runnable Ollama tags: gpt-oss 20B/120B, Qwen3 30B-A3B/Coder, DeepSeek-R1 32B, Mistral Small, GLM-4.7-Flash, Llama 3.3 70B) and a **third-party OSS provider** bucket (BYOK: gpt-oss/Llama via Groq, DeepSeek-V4-Flash, GLM-4.6, Kimi K2.6, Qwen3-Coder via OpenRouter). Picking one fills the engine-correct slug (OpenCode `local/<tag>`, Crush/pi `ollama/<tag>`; provider slugs identical across engines) and rebuilds the command. Slugs are transcribed from a verified catalog — bleeding-edge frontier models are intentionally left out of code defaults. The Add-Agent help line and **Settings → AI Engines** local-setup area now hyperlink two how-to guides (run on open models · set up on a Mac Mini) (`src/shared/ossModels.ts`, `AddAgentModal.tsx`, `AiEnginesSettings.tsx`).
- **Crush no longer dies with `Unknown command` on spawn (the hive protocol now reaches it).** A Crush worker was launched as `crush --model <m> --yolo "You are …(the whole hive protocol)"` — the protocol passed as a positional arg. But bare `crush` is an interactive Bubble Tea TUI on a Cobra root command, which reads the first positional as a **subcommand**, so it aborted with `unknown command "You are…"`; the protocol never reached the model, the worker never learned it was a hive agent, and the PTY died. Crush has no `--prompt` flag and `crush run` is one-shot, so the protocol is now **typed into the TUI** instead: a new preset capability `seedDelivery:'type-into-tui'` makes the spawn drop the positional (`crush [--model m] [--yolo]`) and hand the protocol back as a `seedPrompt`, which the renderer types in as the worker's first turn after a boot-grace — through the **same per-pty write-chain as the inbox-wake nudge**, so the seed and a nudge can never jam onto one line. Covers fresh Crush spawns, restores, and Crush-as-Michael (`src/shared/agentProvider.ts`, `src/main/hive.ts`, `src/main/index.ts`, `src/preload/index.ts`, `src/renderer/src/hooks/useHive.ts`, `AddAgentModal.tsx`, `AgentStrip.tsx`, `store.ts`).
- **Auto restart-and-continue after a first-time engine-CLI install (no dead-end).** When an agent's engine binary (OpenCode/Crush/pi.dev/Codex/…) wasn't installed, the missing-CLI short-circuit ran the provider's installer in the PTY, then printed *"click restart & continue to launch the agent"* — but no such button exists for a not-yet-started agent, so the PTY just sat at `process exited (code 0)` and the agent dead-ended. Now, on a **clean install exit**, the PTY-exit handler auto restart-and-continues: it re-runs the *same* spawn into the *same* pty/window (carrying a `noAutoInstall` flag) so the freshly-installed CLI launches with no user click, and the renderer re-arms that terminal in place (clears the "process exited" line, re-enables input) via a new `pty:relaunch` signal. Provider-agnostic (every engine's installer path) and idempotent by construction — `noAutoInstall` guarantees the installer can never fire twice, and providers with no bundled installer (manual-hint-only) are never armed for relaunch. The install banner copy is now honest ("Installed — launching the agent…") (`src/main/index.ts`, `src/main/pty.ts`, `src/preload/index.ts`, `src/renderer/src/components/terminalPool.ts`).

## [0.3.0] — 2026-06-21

A platform release: the floor stops being Claude-shaped. **Selectable agent engines** make
every hire — and Michael himself — a pluggable engine (Claude Code / Antigravity / Codex /
local providers), each with its own **per-hire skills + MCP catalog** behind a consent UI. A
new **integrations registry + loopback secret broker** turns "connect a service" into a
write-only, registry-driven Settings flow. Michael can now **spawn an ephemeral worker straight
from Slack** — reply, then tear it down safely with worktree GC and token caps — surfaced in a
new **Workers tab**. Plus **temporal date-range skills** and a **worker capability catalog**, a
**Provider / Hive picker** in onboarding and add-agent, the **Agent Gallery** (the rebranded
Hiring Fair) with **six off-the-shelf hires**, feature-aware onboarding, and wake-reliability
hardening. Everything from v0.2.8 and earlier is included.

### Added
- **Selectable agent engines + per-hire capabilities.** A new engine abstraction (`agentProvider` + an `mcpCatalog`, mirrored across a 3-file config) makes the runtime behind each agent *pluggable* — Claude Code, Antigravity, Codex, or a **local provider** (a claw/qwen backend proxy bridge with default-MCP merge). Each hire carries its own **manifest** of allowed skills + MCP servers (a default-deny allowlist over the catalog), with **bundled skills** shipped via Electron `extraResources` (`resources/skills` → `<resources>/skills`) and a **consent UI** that surfaces every skill/MCP a hire wants before it can use it — untrusted hire input is reviewed, never auto-granted.
- **Swappable Michael (god) engine.** The orchestrator is no longer hard-wired to one CLI: `useHive` gains an engine-spawn path, Onboarding gains an **engine picker** for Michael, and a **change-engine flow** lets you re-home the god orchestrator onto a different engine without rebuilding the floor.
- **Integrations registry + loopback secret broker.** A declarative **integrations registry** (`src/shared/integrations.ts`) plus a **loopback secret broker** (`src/main/integrationBroker.ts`): secrets are **write-only** (set once, never read back into the renderer) and reached only through the broker over loopback. A **registry-driven Settings UI** (`IntegrationsRegistry`) renders each integration's config form from the spec — conformed to registry spec v1 — and a first wave of **declarative templates** (the canonical schema + initial YC-style templates) ships in the registry. The `integrations:*` surface is exposed to the renderer through a dedicated preload bridge.
- **God-triggered ephemeral Slack worker loop.** Michael can now **spawn an isolated worker directly in response to a Slack request** — the worker does the work, posts its reply back into the thread, and is then **torn down safely**. Lifecycle hardening adds **worktree garbage collection**, **token-cap wiring** per spawned worker, and a **teardown-safety gate** that refuses to auto-discard a worker's *unintegrated* work. The `pty:spawn` IPC handler was refactored into a reusable `spawnAgentCore` that underpins worker spawning, and a new **Workers tab** surfaces live ephemeral workers in the UI.
- **Temporal date-range skills + worker capability catalog.** A family of date-range skills (`today` / `yesterday` / `thisWeek` / `lastWeek` / `thisMonth` / `thisQuarter` / `thisYear` / `lastMonth` / `lastQuarter` / `lastYear` / `last7Days` / `last30Days` … plus an arbitrary-range `temporal` resolver backed by `temporal/when.mjs`) resolve a named window to concrete ISO dates without hand-math. A **worker capability catalog** lets each spawned worker read exactly which skills and brokered integrations it has and how to call them.
- **Provider / Hive picker UI.** A new `HivePicker` component plus a `ProviderLogo` set (real provider logos) appear in **onboarding** and the **add-agent** flow, so choosing the engine/provider for a hire is a first-class, visual step instead of a free-text command.
- **Agent Gallery + six off-the-shelf hires.** The community gallery is rebranded from *The Hiring Fair* to the **Agent Gallery**, and ships **six ready-made, off-the-shelf hires** you can browse, review, and spawn.
- **Feature-aware onboarding + a permissions & reliability step.** First-run onboarding now adapts to the features you have available and adds an explicit permissions & reliability step.
- **Visible engine-CLI installer.** When the engine binary for a chosen provider is missing, the installer now runs **visibly** instead of failing silently, so a first-time setup self-heals.

### Changed
- **The Hiring Fair → Agent Gallery.** The gallery is renamed throughout (landing page, in-app links, copy) to *Agent Gallery*; existing hire links and the `/hires/` path continue to work.
- **Add-Agent config IA rework + Command Center UX fixes.** The Add-Agent modal's configuration is reorganized around the new engine/capability model, with assorted Command Center UX cleanups.
- **VDE prototype (experimental).** An experimental Virtual Desktop Environment prototype lands behind the scenes, with a Groq chat-completion module (`src/main/groq.ts`) powering its AI assist.

### Fixed
- **Orchestrator delegates opportunistically to existing agents.** Michael now checks the live roster (active agents in `registry.json` + their state in `fleet.json`) before spawning, and prefers routing a task to an existing agent that already fits — above all when the request names one ("ask Pam…", "have Jim…") — instead of reflexively creating a new agent; he only spawns a fresh one when no existing agent is a sensible fit, and says that he checked. Encoded in both the floor orchestrator prompt and the Slack autonomous-request protocol (`src/main/hive.ts`, `src/main/index.ts`).
- **Auto-revive wedged terminals on wake.** A terminal that wedged while the machine slept is now detected and auto-revived when the machine wakes, instead of sitting dead until manually restarted.
- **Catch up missed schedules on wake (power keep-alive hardening).** Scheduled missions whose fire time elapsed while the machine was asleep are now caught up on wake rather than silently skipped (verified by `scripts/verify-keepalive-catchup.mjs`).
- **Worker stale-done guard.** A worker is now released only on a `done` authored *after* it was spawned, so a stale `done` from a prior life can no longer prematurely release or tear down a live worker.
- **Floor: "add agent" button stays on one line.** The add-agent button no longer wraps as the roster fills.
- **Voice button disabled with a tooltip when the Groq key is missing**, instead of failing on click.

### Security
- **Confined the `integrations:test` probe path.** The connectivity-test path for integrations is now constrained so it can't be turned into a secret-exfiltration or SSRF primitive — a brokered, bounded probe rather than an arbitrary outbound request driven by registry/secret input.

## [0.2.8] — 2026-06-15

A feature release: **shareable hires** — package a role-configured agent as a portable
manifest, share it as a file or host it in a gallery, and import it into any office with one
click. Plus **The Hiring Fair**, a community gallery of ready-made roles, and a hardened,
untrusted-input import pipeline.

### Added
- **Shareable hires (#70, #71).** A portable `munder-difflin/hire@1` JSON manifest describing a role-configured agent — name, sprite, provider, model, command flags, goal, capability tags, token budget. Two import paths, one pipeline: a `munderdifflin://hire?src=<https-manifest-url>` deep link (fetched and validated in the main process, queued, then pulled by the renderer on mount) and an *import hire…* button in the Add-Agent modal that reads a local manifest file. Either way the manifest only **pre-fills** the Add-Agent modal behind an "imported" banner; spawning stays an explicit human click — import never auto-spawns. Protocol registration ships for all three platforms (macOS `open-url`, Windows/Linux single-instance lock + cold-start argv forwarding), and packaged builds register the scheme via `electron-builder.yml`.
- **The Hiring Fair — community gallery** at [munderdiffl.in/hires](https://munderdiffl.in/hires/) (`docs/hires/`, static, no build step, served by the existing GitHub Pages setup). Seed roles drawn from the cast (Pam writes docs, Dwight enforces QA, Jim reviews PRs, Creed audits security, Angela audits the office's own token spend, Stanley does the migrations nobody wants), each with a Claude Code / Antigravity / Codex provider toggle (per-provider variants generated from one base manifest), function filters matching the landing page, and a client-side validator identical to the app's alongside a JSON schema (`docs/hires/spec/`). Model suggestions are data-driven (`docs/hires/models.json`), so new models are a one-line update.

### Security
- **A hire manifest is untrusted input — defense in depth.** No auto-spawn and no executable field: `provider: "custom"` is rejected and the binary always comes from the user's local provider preset. Embedded CLI flags are gated by a **default-deny allowlist** (`SAFE_FLAG_NAMES`) — only known-harmless flags pass, nothing system-prompt/settings-related — replacing an earlier denylist that drifted as each CLI added flags. `model` is constrained to a safe charset (`MODEL_RE`), and a command-line quoter neutralizes `cmd.exe` metacharacters (`& | ^ < > ( ) % !`) on **every** spawn path — closing a Windows command-injection class (PoC `"model":"x&calc"`). The manifest fetch is https-only, manual-redirect with per-hop re-validation (kills redirect SSRF into `127.0.0.1` / `169.254.169.254`, including an IPv6-bracket bypass), streamed with a 64 KB byte cap (no trusting `content-length`), a 10s timeout, and ≤5 hops. The dependency-free validator (`src/shared/hire.ts`) is shared by the main process, the renderer, the gallery (`docs/hires/validator.js`), and the JSON schema, so all four stay in sync.

## [0.2.7] — 2026-06-13

A feature release: talk to your agents with your voice, an opt-in enterprise Knowledge
Graph, multi-window "floors", a richer message composer with file/image attachments, the
groundwork for TV-show office themes, and a redesigned landing page — plus composer and
fullscreen polish.

### Added
- **Free Flow voice dictation → message queue (now on by default).** Hold Option to talk; your speech is transcribed by Groq Whisper (`whisper-large-v3-turbo`) straight into the message composer. Gated on a Groq API key, which is encrypted at rest.
- **Enterprise Knowledge Graph v1 (now on by default).** A multimodal store of your own documents / policies / business context, with a CLI agents can query for ranked passages and full documents — so company-specific facts come from your data instead of guesses.
- **Multi-window "floors" (now on by default).** Open isolated office windows, each with its own set of agents and per-PTY routing.
- **Rich message composer — file & image attachments.** Attach files/images (via a "files" button or paste-to-attach), shown as removable chips above a taller, resizable input; you can send with attachments alone.
- **Restore agent sessions across restart, with Restart & Continue (#78).** Agents reattach their prior Claude conversation after an app restart: Michael resumes his session (the orientation prompt is skipped on a genuine resume), and a restored worker re-enters its *existing* worktree instead of re-isolating, so uncommitted work isn't lost. The recorded session transcript is seeded into the target cwd before `--resume` attaches (and `--resume` is only used when the transcript is actually present, so there are no broken resumes against a missing id), and the pooled terminal soft-resets in place — staying live and typeable across a model change or respawn, redrawn at its real fit-derived grid. A per-agent **Restart & Continue** button respawns the session on the same model with resume to redraw a garbled terminal, and Add Agent gains a "resume session" field that reattaches by session id (auto-filling the folder, falling back to a fresh session if the id isn't found).
- **Drag a file onto a terminal to inject its path (#79).** Dropping a file (an image, etc.) onto an agent's terminal now writes its absolute, shell-escaped path into the session — so Claude Code detects the image path in the prompt and attaches it — instead of Electron navigating to the dropped `file://` URL. Backed by `webUtils.getPathForFile` exposed from preload (Electron 32 removed `File.path`); only file drags are intercepted, so text/selection drags still fall through to xterm.
- **TV-show office themes — infrastructure (behind a flag, off by default).** A theme abstraction (`ThemeConfig` + registry/loader), a Settings theme picker with a destructive switch-flow, and the first themed map (Brooklyn-99 precinct). Ships dark via the `tvShowOffices` flag while the remaining maps land.
- **Live GitHub star count** next to the Star buttons on the landing page.

### Changed
- **Composer redesign.** A full-width input above a single tidy control bar (Delegate · Attach · voice · Send) — no dead space from a stacked side column.
- **Landing page redesign.** Bento layout for the #features and #why sections with new SVG illustrations; the #claude section refreshed for v0.2.7.

### Fixed
- **Fullscreen tab bar no longer clipped.** The fullscreen terminal's tab bar is un-clipped.
- **Slack double-ack (#).** A single Slack message delivered as both `app_mention` and `message.*` is now de-duplicated by `channel:ts`, so it's handled exactly once.

## [0.2.6] — 2026-06-10

A polish + reliability patch: the agent terminal renders correctly the moment it opens,
`npm run dev` no longer crashes on a missing sidecar, a Windows ConPTY crash is guarded,
the wall clock becomes a clickable closing-time control, the ASK ME board reads in the
memory font, and the Slack file download is host-pinned.

### Added
- **The office clock is interactive (#64).** The clock on the wall reads the real time, and clicking it opens the closing-time (graceful shutdown) flow.

### Fixed
- **Terminal no longer renders oversized/clipped when an agent boots.** xterm used to fit before its container had a real size and cached the character-cell metrics from before the web font (VT323) loaded, so the welcome banner overflowed and was clipped until you manually resized. The view now waits for a real size, re-measures and re-rasters the WebGL glyph atlas after the font loads, and lets the `ResizeObserver` drive the first fit — so it fits immediately on boot.
- **`npm run dev` no longer crashes on the missing Slack-trigger sidecar (#67).** The `.cjs` sidecar copy now runs as a vite `writeBundle` plugin, so both `dev` and `build` emit `out/main/slack-trigger.cjs` from one place. (v0.2.5 fixed only the packaged-build path, so a fresh clone's `npm run dev` still died at boot.)
- **Windows: node-pty ConPTY `AttachConsole` crash guarded (#65).** Companion to the Antigravity provider work — the main process no longer crashes when ConPTY fails to attach a console.
- **ASK ME reads in the memory font (#63).** The ASK ME board now uses VT323 instead of the chunky Pixelify Sans, matching the rest of the memory surfaces.

### Security
- **`downloadSlackFile()` host-pinned to Slack.** The Slack bot token is now only ever sent to `slack.com` / `*.slack.com` hosts — a defense-in-depth guard before the `Authorization: Bearer` header is attached (the URL is already Slack-issued + HMAC-verified, so this hardens against a future redirect/parsing change).

## [0.2.5] — 2026-06-10

A reliability + reach patch: a Windows-terminal regression fix, an agent-lifecycle
cleanup that ends the breaker inbox-flood, Slack requests that actually reply with
substance, a delegate toggle, six new tutorials/blogs, and an enriched landing diagram.

### Added
- **Delegate-to-agents toggle.** A toggle switch above the Send button in the god orchestrator's composer prepends a delegation instruction so a request fans out to available agents (and is handled one-by-one if none are free). God-only, default off.
- **AUTONOMOUS REQUEST PROTOCOL for Slack-origin requests.** Inbound Slack requests now run fully autonomously: god routes the request to the most-relevant agent, that agent does the work and **posts its substantive result back into the Slack thread itself**, then reports to god. It pauses only for high-severity actions (pushing to main, spawning infrastructure/paid services, deleting files it didn't create), and any decision it needs is asked as a numbered-options reply in the thread and correlated back by `thread_ts`.
- **Six new tutorials & blog posts** — webhook setup, the full Slack setup, deploying an automated PR-reviewer agent, deploying a blog-writer agent, why CLI agents are so powerful (and how the hive cuts token use), and why a mixed-capability swarm beats a clone army.
- **Enriched "how it works" landing diagram** — Slack / Webhook / Schedule triggers feeding the orchestrator, plus a band showing each agent in its own isolated local git worktree.

### Fixed
- **Windows agent terminals no longer die on Program-Files installs (#55).** `cmd.exe` is now invoked with a properly double-quoted command line (`/d /s /c`), so a Claude/node path containing a space (e.g. `C:\Program Files\…`) launches instead of splitting on the space.
- **Orphaned-agent lifecycle cluster (#56/#57/#58).** A stale agent entry with no live terminal no longer (a) re-writes a frozen cost-ledger row every ~30s, (b) trips the circuit breaker into an unclearable inbox-flood to the orchestrator, or (c) lingers un-archived — a startup migration archives no-PTY entries and the breaker now skips assistant/orphaned shells.
- **Slack replies are real answers, not empty confirmations.** Worker agents post the actual outcome into the thread; the orchestrator's auto-summary is now a fallback that never posts a bare "✅" with no content and skips threads already answered directly.

### Removed
- **Reverted the unfinished compact-protocol feature** — it was only half-wired (main-side committed, renderer-side unfinished) and broke the web typecheck. It will return fully wired in a later release.

## [0.2.4] — 2026-06-09

A multi-provider patch: Codex graduates to **full hive parity** via a native
lifecycle-hook bridge, the god orchestrator opens to its terminal by default, and a
handful of resilience fixes land.

### Added
- **Codex lifecycle-hook bridge — full hive parity.** Codex now joins the hive as a first-class, hive-aware provider: a native lifecycle-hook bridge maps Codex's events into the existing hook pipeline (live status + inbox-drain + outbox routing), and agy/codex dispatch is unified behind one path. Verified running hive-aware in bypass-permissions mode. (#47, #54)
- **Codex hook discovery via `config.toml [hooks]`.** The bridge registers through Codex's `config.toml [hooks]` surface rather than a bare `hooks.json`, matching how Codex actually discovers lifecycle hooks.

### Changed
- **God orchestrator opens to the Terminal sidebar by default.** Selecting the god agent no longer reopens a stale "ASK ME" tab — a leftover command-center tab request is cleared on select, so the panel mounts to its terminal default. The ASK ME tab is still one click away.
- **Landing + blog refreshed for multi-provider.** The landing page now presents Claude Code, Antigravity (Gemini), and OpenAI Codex as equal first-class providers (with a one-line mobile-friendly badge), and a grand v0.2.4 launch post + technical walkthrough replace the v0.2.3 posts.

### Fixed
- **Slack/webhook tunnel no longer crashes at load.** `tunnelmole` is ESM-only; a static `import` in the CommonJS-bundled Electron main process threw `ERR_REQUIRE_ESM`. It's now loaded via a dynamic `import()` inside `openTunnel()`, so the public ingress actually starts.
- **Heartbeat re-engages the god on an unread actionable inbox** — not only when the floor is quiet — so worker/human mail is drained promptly.
- **Slack done-summary stops retrying on terminal errors.** A permanently-failing post (e.g. the bot token missing `chat:write` → `missing_scope`) is now recorded and logged once instead of retrying every 5s and flooding the console; transient errors still retry.

## [0.2.3] — 2026-06-09

A multi-provider release: the floor is no longer Claude-only. Antigravity (Gemini)
and Codex agents become first-class hive participants, schedules get their own tab,
and the Slack / webhook ingress is moved off the flaky public tunnel.

### Added
- **First-class Antigravity (Gemini / `agy`) provider.** A worker can now run the Antigravity CLI as a full hive participant. Because `agy` has no Claude-style `--append-system-prompt`/`--settings` hooks, the hive identity + protocol ride in as the session's initial prompt, and a native `agy-hook` bridge normalizes Antigravity's lifecycle events into the existing hook pipeline so a Gemini worker gets the same live status + inbox-drain as Claude — on the subscription, no API key. (#54)
- **Schedules tab.** Recurring auto-dispatched missions (and the adaptive heartbeat) get their own Command-Center tab instead of an inline section. (#50)
- **Terminal work-order handoff for hookless providers.** A provider with no inbox-drain path now receives hive mail as a `WORK ORDER FROM HIVE` typed into its terminal, falling back to a god-bounce only if the renderer is unavailable. (#53)

### Fixed
- **Codex agents now follow the hive protocol and message back.** Codex spawned without the hive protocol or any hook, so it never read its inbox or wrote its outbox. Codex is now a non-hive-aware-but-inbox-capable provider: the protocol is injected as its initial (positional) prompt, its outbox is drained provider-agnostically by the router, and inbox mail reaches it via the renderer's idle inbox-wake nudge. Codex and Antigravity coexist in the provider union. (#47, #54)
- **Slack + webhook public URL no longer silently breaks.** The ingress used `localtunnel`/loca.lt, which now serves a browser interstitial that fails Slack's `url_verification` POST and breaks saved webhook URLs. Both `slack.ts` and `webhook.ts` now use **tunnelmole** (MIT, POSTs pass straight through), and a failed tunnel surfaces a real error instead of a silent "started" with no URL.

## [0.2.2] — 2026-06-07

A community polish release — almost entirely the work of @Gulum: a live context-window
gauge on every agent card, sharper terminals, correct Windows metering, and dispatch that
always routes through the god.

### Added
- **Live context-window gauge on each agent card.** A Claude Code statusLine pushes the session's exact token count and real context-window size after every response, so each agent card shows a precise live fuel gauge drawn from Claude Code itself instead of a transcript estimate. The gauge also zeroes the instant you send `/clear`, rather than briefly showing the previous session's full bar until the next response. (Thanks @Gulum — #12, closes #11.)
- **Per-session terminal theme toggle + Unicode 11 emoji widths.** Each terminal session can now switch its Claude theme independently, and emoji column widths follow Unicode 11 so wide glyphs stop nudging the cursor out of alignment. (The WebGL renderer, copy/paste, and `minimumContrastRatio` from v0.2.0 are kept as-is.) (Thanks @Gulum — #26.)
- **All human dispatch flows through the god.** Every Command Center dispatch now mails the god (`Task from the human`) instead of writing straight into a worker's inbox; the worker dropdown becomes a **suggested owner** (Michael still decides), so nothing skips the orchestrator. (Thanks @Gulum — #45, fixes #44.)
- **Dedicated context-window row on the monitor tab.** The Floor monitor's cumulative budget bar was being misread as a per-agent context gauge; a separate `ctx` context-window row now sits alongside it, so the live context window and the cumulative budget aren't confused. (Thanks @Gulum — #46.)

### Fixed
- **Windows usage meter no longer reads 0/0.** The transcript reconciler built the per-project directory name with the POSIX rule, but Claude Code on Windows encodes *every* non-alphanumeric character (including the drive colon), so the meter never found the transcript and always read 0 tokens / $0.00. (Thanks @Gulum — #34, fixes #10.)
- **Send-only assistant mail no longer black-holes.** Direct mail to the send-only prep assistant landed in an inbox nothing reads; the router now bounces it to the god (subject prefixed `[bounced …]`) instead of dropping it. (Thanks @Gulum — #33, fixes #32.)
- **Boot banner no longer stacks in scrollback.** `tryFit()` fired `resizePty` on every fit even when the dimensions were unchanged; redundant resizes are now skipped, so the boot banner stops re-stacking in the terminal history. (Thanks @Gulum — #8.)
- **Visible text-select cursor on the cream theme.** The hovering I-beam (an OS cursor that CSS color hints can't touch, drawn white by several Windows schemes) is now an inked I-beam with a halo, so it stays visible over the light terminal. (Thanks @Gulum — #39.)

### Acknowledgements
This release is almost entirely the work of @Gulum — thank you. Maintained by @chaitanyagiri.

## [0.2.1] — 2026-06-07

A small follow-up to v0.2.0 that makes the scheduler considerate of agents that are mid-task.

### Changed
- **Scheduled auto-compaction is queued, not forced.** The hourly ops-standup's terminal compaction is now enqueued per agent and delivered only when that agent is idle (deduped — at most one `/compact` pending at a time), so it compacts *between* steps instead of jamming a working terminal mid-step. The standup prompt now asks each agent to summarise its current task and next step, then resume from the same point after compacting.
- **Heartbeat is inbox-driven.** The floor heartbeat (`reengageGod`) no longer types directly into Michael's terminal; it drops its digest in his inbox, which the busy-aware inbox-wake delivers once he's idle.

### Docs
- Expanded the README roadmap (chat integrations, pluggable agent CLIs, realtime Michael).

## [0.2.0] — 2026-06-07

The observability and control release. v0.2.0 makes the fleet visible and keeps it in check — and it's a community release in the most literal sense: most of the work below came from external contributors. Huge thanks to everyone credited.

### Added
- **Command Center overhaul.** Michael's control surface was reworked into the place you actually run the floor from — the roster, dispatch, schedules, memory, and activity views now carry the new live signals (token budgets, telemetry, breaker state) without becoming a wall of numbers.
- **Per-agent token budgets + live fleet monitoring.** Every agent carries a token budget, and the floor monitors consumption live so a single agent can't quietly run the bill up.
- **Live OTel telemetry collector + per-model cost.** A built-in OpenTelemetry collector and a `UsageProvider` seam feed real usage in, with per-model cost attribution (interim transcript-backed stub behind the seam to start).
- **Fleet grid + per-agent tool-span waterfall.** A live grid of the whole fleet, plus a per-agent tool-span waterfall that shows what an agent spent its turn doing — which tool calls ran, in what order, for how long.
- **Agent-card context-window gauge.** The agent card's progress bar is repurposed into a context-window gauge so you can see at a glance how close each agent is to filling its context. (Thanks @Gulum — #12.)
- **Circuit breaker.** A steer → constrain → stop ladder plus a cost/runaway guard, fed by hook signals (repeated identical tool calls), an `onApiError` seam for error-storm trips, and budget config.
- **Scheduler heartbeat.** A heartbeat beat tracks each agent's last output (quiet/idle signals), enforces circuit-breaker policy, and adds spawn guardrails; the SCHEDULES view shows the heartbeat row plus last-fired / next-fired times. (Thanks @albozes — #2.)
- **Human-in-the-loop, mid-run.** A HITL gate, mid-run steer, and graceful stop all delivered through hook returns — approve, redirect, or cleanly halt an agent mid-turn instead of yanking it.
- **Durable SQLite persistence (Phase A).** A SQLite durable store persists window bounds and history, alongside a persisted `session_id` and a durable cost ledger (`cost-ledger.jsonl`) so cost and provenance survive restarts.
- **MemoryReflector — memory condensation.** The janitor's missing condense half: a reflector that condenses memory instead of only mining it, keeping the semantic store lean.
- **Configurable hive/memory home folder.** Point the hive and memory home at a folder of your choosing, with a safe move that relocates existing data.
- **One-click "Restore team."** After a harness restart, a single click brings back the last session's workers — no more re-adding agents one by one. (Thanks @Gulum — #16.)
- **Delete scheduled missions.** Scheduled missions now have a delete button. (Thanks @Gulum — #9.)
- **New avatar states.** A compacting state (on `PreCompact`) and a looping state (when the breaker engages) so the floor reflects what the control layer is doing.

### Fixed
- **Terminal contrast + HiDPI legibility.** A `minimumContrastRatio` floor and a tuned light palette (including dual fg/bg-legible green/yellow) keep text legible on coloured backgrounds across both the new and legacy terminal views.
- **Crisp, readable floor text.** A HiDPI canvas, bold bubbles, and a walk-flicker fix sharpen office-floor text; thought-cloud text now stays 1:1 when the window shrinks. (Thanks @Gulum — #20.)
- **Terminal no longer jumps to the top of history on first scroll**, and the viewport dead zone is gone (re-sync routed through xterm, not the DOM). (Thanks @Gulum — #8.)
- **Windows: keep the hive running behind the lock screen.** The hive no longer freezes when Windows locks — keep-awake plus no throttling. (Thanks @Gulum — #18.)
- **Live agent statuses** and **composer-draft fixes** for the message-queue composer. (Thanks @Gulum — #7, #27, #28.)
- **Palace writer-lock serialization** and **Windows named-pipe + mempalace detection** so the hook server and semantic memory work on Windows. (Thanks @Xileck.)
- **Per-PTY input serialization** so the boot sequence can't jam mid-spawn; restored `+x` on the `node-pty` spawn-helper so agents can spawn.
- **GOD orchestration tabs** are now scrollable; the title-bar settings button is a clear gear chip.
- **Global `defaultModel` wins over role tier** (an explicit per-agent pick still wins); cost-ledger row is fully snake_case for a 1:1 SQLite migration (#4).

### Acknowledgements
Reported / requested by the community: @JLAD75 (Windows hive router / `hooks.sock` — #1), @billrehm (Windows GOD-spawn error 193 — #22), @darrensheffield (uv-not-installed assumption — #30; macOS Gatekeeper — #29), @pdurlej (first-class Codex CLI provider request — #21), @wild-gobatz (agents showing idle until clicked — #3). Maintained by @chaitanyagiri.

## [0.1.9] — 2026-06-06

### Added
- **Hourly ops standup.** A built-in scheduled mission (enabled by default) where the GOD orchestrator reviews every agent — who's doing what, whether tasks are on track, and whether agents are still running — and **compacts each terminal's context** on the same hourly cadence to keep agents lean. Toggle it in the Command Center; a one-time migration seeds it into existing installs (and won't re-add it once deleted).

### Fixed
- **Agents exited on their own at a "Bypass Permissions mode" prompt.** Agents spawn with `--permission-mode bypassPermissions`, which on a fresh machine shows a one-time interactive "WARNING: Bypass Permissions mode … 1. No, exit / 2. Yes, I accept" prompt the terminal couldn't answer, so the agent exited code 1 within seconds. The harness now idempotently pre-accepts Claude Code's dangerous-mode warning (`skipDangerousModePermissionPrompt` / `skipAutoPermissionPrompt`) and per-folder trust before each spawn.
- **Blog cards.** The colored thumbnail tile sat flush against each card's bold border; it's now inset with padding for breathing room (desktop + mobile).
- **Windows: hook server + semantic memory.** The hook server now binds a named pipe on Windows (where Node IPC isn't a filesystem socket), and `mempalace` is detected via `where` + the standard `.exe` install locations so semantic memory works on Windows. POSIX behavior unchanged. (Thanks @Xileck — #4.)
- **Palace mining writer-lock collisions.** Mining is now serialized — a single writer at a time with a re-entrancy guard — fixing "palace is held by PID …" failures when multiple agents mined concurrently. (Thanks @Xileck — #5.)
- **MemPalace index noise.** Each agent's `.gitignore` is ensured before mining so `settings.json`, the cursor file, and raw inbox/outbox message JSON stay out of the semantic index — keeping recall and wake-up focused on real memory.

## [0.1.8] — 2026-06-05

### Fixed
- **Windows agent spawn.** Launching an agent on Windows failed with `cannot create process, error code: 2` (ENOENT) because binary and PATH resolution were Unix-only (`SHELL`/`/bin/zsh`, `-ilc`, `which`, and Unix-only fallback paths). Windows now resolves `claude` via `where`, checks the standard Windows install locations (`%APPDATA%\npm\claude.cmd`, `%LOCALAPPDATA%\Programs\claude`, `%USERPROFILE%\.claude\local`), uses the process `PATH` directly (no login-shell probe), and recognizes Windows-style (`\`) absolute paths. macOS and Linux resolution is unchanged; the Unix fallbacks now also include `~/.volta/bin`.

## [0.1.7] — 2026-06-04

### Added
- **Slack → Michael's queue.** A new Slack integration (Settings → Slack) pipes a channel's messages straight into Michael's message queue — paste a message in Slack and it lands in his queue exactly as if you'd typed it. Off by default; every request is verified with your Slack signing secret (HMAC + 5-minute replay guard) before it's accepted, and a localtunnel exposes the local webhook for Slack's Event Subscriptions.

### Changed
- **Approvals are now native.** The in-app approvals queue/panel is removed in favor of native Claude Code human-in-the-loop prompts. A `to:"human"` decision now reaches you through Michael's session and native permission prompts — approvable from your phone via `/remote-control` — and Michael boots straight into running the floor.

### Fixed
- The floating approvals panel could re-queue an item when you approved it (`resolveApproval` re-routed the message back into the queue). Moving to native HITL removes the panel and the bug.

## [0.1.6] — 2026-06-04

### Added
- **Per-agent git worktrees.** A 'Git isolation' toggle in Add Agent auto-provisions a dedicated worktree (`<harnessHome>/worktrees/<agentId>/`) on spawn and tears it down on kill. Agents on the same repo never collide on branches.
- **Task kanban with dependencies.** A Tasks tab in the Command Center renders a full kanban board (todo / doing / blocked / done). Each task carries an assignee, a `dependsOn[]` list, priority, and description — and persists in `hive/tasks.json` via a new `hive:writeTasks` IPC channel.
- **Scheduled missions.** A Schedules section in the Floor tab lets you define recurring auto-dispatch missions (label, interval, target agent, body). The main process fires each on a `setInterval`, stamps `lastFiredAt`, and persists the list in config.
- **Real token & cost telemetry.** The Activity tab reads `~/.claude/projects/` JSONL transcripts — the same files Claude Code writes — and displays actual input/output/cache token counts and estimated USD cost per agent per model. No more proxy tool-call counts.
- **Global hive text search.** Full-text search across `board.md`, `tasks.json`, and all agent `memory.md` files, available in the Memory tab alongside MemPalace semantic search.
- **Threaded chat.** A Messages tab in each agent's sidebar renders every hive message grouped by conversation with full reply chains and an inline reply form.
- **Memory graph.** A visual graph in the Command Center Memory tab maps agents and their knowledge relationships.
- **GitHub issue ingestion.** An Issues section in the Floor tab pulls open issues from any registered repo via `gh issue list` and lets you assign them to any agent with one click.
- **CI status watcher.** A CI Status section in the Activity tab polls `gh run list` for every registered repo and shows live pass/fail/in-progress status for GitHub Actions runs.
- **Desktop notifications.** Native OS notifications fire when an agent finishes a task or is waiting for your input. Toggle in Settings.
- **Agent archival.** Closing an agent's tab archives it (memory + history intact) rather than deleting it permanently.

### Fixed
- Scheduler now honors `lastFiredAt` on config reload — missions don't double-fire after a save.
- PTY lifecycle teardown runs on natural process exit as well as explicit kill, so worktrees are cleaned up reliably.
- Task IDs fall back to a stable UUID when the title is empty; `writeTasks` IPC validates its input.

## [0.1.5] — 2026-06-04

### Added
- **Dwight, Michael's prep assistant.** A persistent, visible assistant agent
  (Sonnet, 1M context) spawns on startup. A global **enrich** toggle routes
  Michael's queued prompts through Dwight first — he gathers repo context and
  rewrites the prompt, then forwards it to Michael through the hive; toggle it off
  and prompts go straight to Michael.
- **Michael's Command Center.** His sidebar becomes a control surface with
  Terminal, Floor (agent roster + **per-agent model selector** with safe restart,
  a dispatch box, and working dirs), Memory (MemPalace search + per-agent memory),
  and Activity (live log feed + board + usage proxy), plus a copyable Claude
  command handbook.
- **Per-agent model selection** — a model picker in **Add Agent**, a shared model
  list, and a message-queue composer with the enrich toggle.
- **Getting-started tutorial** on the blog (canonical install + first-run walkthrough),
  with Blog/tutorial CTAs and a redesigned "How it works" section on the landing page.

### Fixed
- Agents no longer read **"idle" while still working** — a Stop blocked mid-turn now
  reports `blocked` so the UI keeps the agent in its working state.
- Long agent thought/tool labels now **word-wrap inside their cards** instead of
  overflowing the bubble horizontally (Pixi word-wrap with a raw-length cap so a
  pathological string can't grow a runaway-tall card).
- Switching agent terminals now lands at the **latest output**, while resizes
  preserve scroll position; the idle action label no longer echoes the "idle" badge.

## [0.1.4] — 2026-06-04

### Added
- **Signed macOS builds.** The app now ships with a hardened-runtime Developer ID
  signature (notarization is attempted in CI and stapled when it succeeds; the build is
  best-effort, so a notarization hiccup never blocks a release). Because macOS binds a
  folder-access (TCC) grant to a stable code signature, you're now prompted for
  Documents/Desktop/Downloads access **once** instead of on every agent action.
  Usage-description strings explain each prompt. Signing/notarization run in CI only
  when Apple credentials are present, so contributor builds stay unsigned and green.
- **Blog at [/blog](https://munderdiffl.in/blog/)** — an Eleventy-generated static blog
  sharing the landing page's neo-brutalist design system, seeded with the first posts
  on long-term memory, multi-agent harnesses, and MemPalace, plus tag/topic indexes and
  an RSS feed.
- **On-site SEO/AEO metadata** — JSON-LD, `robots.txt`, a root `sitemap.xml`, and richer
  link-unfurl/meta tags across the site.

## [0.1.3] — 2026-06-01

### Added
- **Settings panel** (title-bar gear) with a **Reset & start over** action that wipes
  Michael's memories, the entire hive (every agent, message, task, and the board), and
  the semantic-memory palace, then relaunches the app into onboarding.
- Boot loader ("clocking in") shown while the GOD agent initializes, so returning users
  no longer see the empty "add agent" screen during startup.

### Fixed
- Crash dialog on quit caused by sending IPC to an already-destroyed window during
  teardown; all renderer sends are now destroyed-safe and shutdown steps are best-effort.
- Michael no longer marches to the door flagged "needs you" right after finishing a turn —
  idle "waiting for input" notifications now let him linger at his desk instead of
  escalating as a blocked/needs-action state.

## [Brand & rename]

### Added
- Brand identity: **Munder Difflin** — logo (`docs/logo.svg`), square mark
  (`docs/logo-mark.svg`), and hero banner (`docs/banner.svg`).
- Landing page at `docs/index.html` (GitHub Pages–ready).
- In-app branding: window title, boot screen, title-bar `MD` badge, and fullscreen
  header captions.
- Open-source community files: `SECURITY.md`, `CHANGELOG.md`, issue/PR templates, and a
  CI workflow.

### Changed
- Renamed the project from *Claude Terminal Harness* to **Munder Difflin** across the
  README, docs (`SPEC.md`, `DESIGN.md`, `HIVE.md`), `package.json`, and the app UI.

## [0.1.0] — 2026

Initial working prototype.

### Added
- Electron + React + TypeScript shell (electron-vite).
- Real terminals via `node-pty`, rendered with xterm.js; multi-agent spawn/write/
  resize/kill over typed IPC (`window.cth`).
- Pixi.js office floor: Tiled map, camera, recolored cast, pathfinding, seat assignment,
  tool bubbles, and message envelopes.
- The hive: on-disk multi-agent layer (`hive.ts`), hook server + `cth-hook` shim and
  `Stop`-loop (`hooks.ts`), and a semantic memory layer (`memory.ts`).
- GOD orchestrator agent, approvals queue, and memory search panel.
- Sandboxed file browser + CodeMirror editor and a git tab (status, log, branches,
  commit graph).
- Onboarding wizard, safe-quit guard, and a tokenized SNES/Animal-Crossing design
  system.

[0.1.0]: https://github.com/chaitanyagiri/munder-difflin/releases/tag/v0.1.0