Merge branch 'main' into cw/account-ui-conflict-1772803526476

# Conflicts:
#	apps/server/trpc/router.test.ts
#	docs/server-api.md

docs/agent.md

@@ -11,7 +11,7 @@
 | `process-manager.ts` | `AgentProcessManager` — worktree creation, command building, detached spawn |
 | `output-handler.ts` | `OutputHandler` — JSONL stream parsing, completion detection, proposal creation, task dedup, task dependency persistence |
 | `file-tailer.ts` | `FileTailer` — watches output files, fires parser + raw content callbacks |
-| `file-io.ts` | Input/output file I/O: frontmatter writing, signal.json reading, tiptap conversion. Output files support `action` field (create/update/delete) for chat mode CRUD. |
+| `file-io.ts` | Input/output file I/O: frontmatter writing, signal.json reading, tiptap conversion. Output files support `action` field (create/update/delete) for chat mode CRUD. Includes `writeErrandManifest()` for errand agent input files. |
 | `markdown-to-tiptap.ts` | Markdown to Tiptap JSON conversion using MarkdownManager |
 | `index.ts` | Public exports, `ClaudeAgentManager` deprecated alias |
 
@@ -24,15 +24,15 @@
 | `accounts/` | Account discovery, config dir setup, credential management, usage API |
 | `credentials/` | `AccountCredentialManager` — credential injection per account |
 | `lifecycle/` | `LifecycleController` — retry policy, signal recovery, missing signal instructions |
-| `prompts/` | Mode-specific prompt builders (execute, discuss, plan, detail, refine, chat) + shared blocks (test integrity, deviation rules, git workflow, session startup, progress tracking) + inter-agent communication instructions |
+| `prompts/` | Mode-specific prompt builders (execute, discuss, plan, detail, refine, chat, conflict-resolution, errand) + shared blocks (test integrity, deviation rules, git workflow, session startup, progress tracking) + inter-agent communication instructions. Conflict-resolution uses a minimal inline startup (pwd, git status, CLAUDE.md) instead of the full `SESSION_STARTUP`/`CONTEXT_MANAGEMENT` blocks. |
 
 ## Key Flows
 
 ### Spawning an Agent
 
 1. **tRPC procedure** calls `agentManager.spawn(options)`
-2. Manager generates alias (adjective-animal), creates DB record
-3. `AgentProcessManager.createWorktree()` — creates git worktree at `.cw-worktrees/agent/<alias>/`
+2. Manager generates alias (adjective-animal), creates DB record. Appends inter-agent communication and preview instructions unless `skipPromptExtras: true` (used by conflict-resolution agents to keep prompts lean).
+3. `AgentProcessManager.createProjectWorktrees()` — creates git worktrees at `agent-workdirs/<alias>/<project>/`. After creation, each project subdirectory is verified to exist; missing worktrees throw immediately to prevent agents running in the wrong directory.
 4. `file-io.writeInputFiles()` — writes `.cw/input/` with assignment files (initiative, pages, phase, task) and read-only context dirs (`context/phases/`, `context/tasks/`)
 5. Provider config builds spawn command via `buildSpawnCommand()`
 6. `spawnDetached()` — launches detached child process with file output redirection
@@ -115,6 +115,30 @@ cw account add --token <token> --email user@example.com
 
 Stored as `credentials: {"claudeAiOauth":{"accessToken":"<token>"}}` and `configJson: {"hasCompletedOnboarding":true}`.
 
+## Errand Agent Support
+
+### `sendUserMessage(agentId, message)`
+
+Delivers a user message directly to a running or idle errand agent without going through the conversations table. Used by the `errand.sendMessage` tRPC procedure.
+
+**Steps**: look up agent → validate status (`running`|`idle`) → validate `sessionId` → clear signal.json → update status to `running` → build resume command → stop active tailer/poll → spawn detached → start polling.
+
+**Key difference from `resumeForConversation`**: no `conversationResumeLocks`, no conversations table entry, raw message passed as resume prompt.
+
+### `writeErrandManifest(options)`
+
+Writes errand input files to `<agentWorkdir>/.cw/input/`:
+
+- `errand.md` — YAML frontmatter with `id`, `description`, `branch`, `project`
+- `manifest.json` — `{ errandId, agentId, agentName, mode: "errand" }` (no `files`/`contextFiles` arrays)
+- `expected-pwd.txt` — the agent workdir path
+
+Written in order: `errand.md` first, `manifest.json` last (same discipline as `writeInputFiles`).
+
+### `buildErrandPrompt(description)`
+
+Builds the initial prompt for errand agents. Exported from `prompts/errand.ts` and re-exported from `prompts/index.ts`. The prompt instructs the agent to make only the changes needed for the description and write `signal.json` when done.
+
 ## Auto-Resume for Conversations
 
 When Agent A asks Agent B a question via `cw ask` and Agent B is idle, the conversation router automatically resumes Agent B's session. This mirrors the `resumeForCommit()` pattern.
@@ -153,9 +177,10 @@ Agent output is persisted to `agent_log_chunks` table and drives all live stream
 - DB insert → `agent:output` event emission (single source of truth for UI)
 - No FK to agents — survives agent deletion
 - Session tracking: spawn=1, resume=previousMax+1
-- Read path (`getAgentOutput` tRPC): concatenates all DB chunks (no file fallback)
-- Live path (`onAgentOutput` subscription): listens for `agent:output` events
-- Frontend: initial query loads from DB, subscription accumulates raw JSONL, both parsed via `parseAgentOutput()`
+- Read path (`getAgentOutput` tRPC): returns timestamped chunks `{ content, createdAt }[]` from DB
+- Live path (`onAgentOutput` subscription): listens for `agent:output` events (client stamps with `Date.now()`)
+- Frontend: initial query loads timestamped chunks, subscription accumulates live chunks, both parsed via `parseAgentOutput()` which accepts `TimestampedChunk[]`
+- Timestamps displayed inline (HH:MM:SS) on text, tool_call, system, and session_end messages
 
 ## Inter-Agent Communication
 
@@ -236,7 +261,7 @@ All prompts follow a consistent tag ordering:
 |------|------|--------------------|
 | **execute** | `execute.ts` | `<task>`, `<execution_protocol>`, `<anti_patterns>`, `<scope_rules>` |
 | **plan** | `plan.ts` | `<phase_design>`, `<dependencies>`, `<file_ownership>`, `<specificity>`, `<existing_context>` |
-| **detail** | `detail.ts` | `<task_body_requirements>`, `<file_ownership>`, `<task_sizing>`, `<checkpoint_tasks>`, `<existing_context>` |
+| **detail** | `detail.ts` | `<task_body_requirements>`, `<file_ownership>`, `<task_sizing>`, `<existing_context>` |
 | **discuss** | `discuss.ts` | `<analysis_method>`, `<question_quality>`, `<decision_quality>`, `<question_categories>`, `<rules>` |
 | **refine** | `refine.ts` | `<improvement_priorities>`, `<rules>` |
 | **chat** | `chat.ts` | `<chat_history>`, `<instruction>` — iterative refinement loop, uses action field (create/update/delete) in output files, signals "questions" after each change to stay alive |

docs/cli-config.md

@@ -123,6 +123,7 @@ All three commands output JSON for programmatic agent consumption.
 | `list` | Show accounts with exhaustion status |
 | `remove <id>` | Remove account |
 | `refresh` | Clear expired exhaustion markers |
+| `extract [--email <email>]` | Extract current Claude credentials as JSON (no server required) |
 
 ## Server Wiring
 

docs/database-migrations.md

@@ -5,7 +5,7 @@ This project uses [drizzle-kit](https://orm.drizzle.team/kit-docs/overview) for
 ## Overview
 
 - **Schema definition:** `apps/server/db/schema.ts` (drizzle-orm table definitions)
-- **Migration output:** `apps/server/drizzle/` directory (SQL files + meta journal)
+- **Migration output:** `apps/server/drizzle/` directory (SQL files + `meta/_journal.json` + `meta/NNNN_snapshot.json`)
 - **Config:** `drizzle.config.ts`
 - **Runtime migrator:** `apps/server/db/ensure-schema.ts` (calls `drizzle-orm/better-sqlite3/migrator`)
 
@@ -13,6 +13,8 @@ This project uses [drizzle-kit](https://orm.drizzle.team/kit-docs/overview) for
 
 On every server startup, `ensureSchema(db)` runs all pending migrations from the `apps/server/drizzle/` folder. Drizzle tracks applied migrations in a `__drizzle_migrations` table so only new migrations are applied. This is safe to call repeatedly.
 
+The migrator discovers migrations via `apps/server/drizzle/meta/_journal.json` — **not** by scanning the filesystem.
+
 ## Workflow
 
 ### Making schema changes
@@ -23,7 +25,17 @@ On every server startup, `ensureSchema(db)` runs all pending migrations from the
    npx drizzle-kit generate
    ```
 3. Review the generated SQL in `apps/server/drizzle/NNNN_*.sql`
-4. Commit the migration file along with your schema change
+4. Verify multi-statement migrations use `--> statement-breakpoint` between statements (required by better-sqlite3 which only allows one statement per `prepare()` call)
+5. Commit the migration file, snapshot, and journal update together
+
+### Important: statement breakpoints
+
+better-sqlite3 rejects SQL with multiple statements in a single `prepare()` call. Drizzle-kit splits on `--> statement-breakpoint`. If you hand-write or edit a migration with multiple statements, append `--> statement-breakpoint` to the end of each statement line (before the next statement):
+
+```sql
+ALTER TABLE foo ADD COLUMN bar TEXT;--> statement-breakpoint
+CREATE INDEX foo_bar_idx ON foo(bar);
+```
 
 ### Applying migrations
 
@@ -31,20 +43,15 @@ Migrations are applied automatically on server startup. No manual step needed.
 
 For tests, the same `ensureSchema()` function is called on in-memory SQLite databases in `apps/server/db/repositories/drizzle/test-helpers.ts`.
 
-### Checking migration status
+## History
 
-```bash
-# See what drizzle-kit would generate (dry run)
-npx drizzle-kit generate --dry-run
-
-# Open drizzle studio to inspect the database
-npx drizzle-kit studio
-```
+Migrations 0000–0007 were generated by `drizzle-kit generate`. Migrations 0008–0032 were hand-written (the snapshots fell out of sync). A schema-derived snapshot was restored at 0032, so `drizzle-kit generate` works normally from that point forward.
 
 ## Rules
 
-- **Never hand-write migration SQL.** Always use `drizzle-kit generate` from the schema.
+- **Use `drizzle-kit generate`** for new migrations. It reads schema.ts, diffs against the last snapshot, and generates both SQL + snapshot automatically.
 - **Never use raw CREATE TABLE statements** for schema initialization. The migration system handles this.
 - **Always commit migration files.** They are the source of truth for database evolution.
 - **Migration files are immutable.** Once committed, never edit them. Make a new migration instead.
-- **Test with `npx vitest run`** after generating migrations to verify they work with in-memory databases.
+- **Keep schema.ts in sync.** The schema file is the source of truth for TypeScript types; migrations are the source of truth for database DDL. Both must reflect the same structure.
+- **Test with `npm test`** after generating migrations to verify they work with in-memory databases.

docs/database.md

@@ -5,8 +5,8 @@
 ## Architecture
 
 - **Schema**: `apps/server/db/schema.ts` — all tables, columns, relations
-- **Ports** (interfaces): `apps/server/db/repositories/*.ts` — 13 repository interfaces
-- **Adapters** (implementations): `apps/server/db/repositories/drizzle/*.ts` — 13 Drizzle adapters
+- **Ports** (interfaces): `apps/server/db/repositories/*.ts` — 14 repository interfaces
+- **Adapters** (implementations): `apps/server/db/repositories/drizzle/*.ts` — 14 Drizzle adapters
 - **Barrel exports**: `apps/server/db/index.ts` re-exports everything
 
 All adapters use nanoid() for IDs, auto-manage timestamps, and use Drizzle's `.returning()` for atomic reads after writes.
@@ -29,7 +29,8 @@ All adapters use nanoid() for IDs, auto-manage timestamps, and use Drizzle's `.r
 | initiativeId | text FK → initiatives (cascade) | |
 | name | text NOT NULL | |
 | content | text nullable | Tiptap JSON |
-| status | text enum | 'pending' \| 'approved' \| 'in_progress' \| 'completed' \| 'blocked' |
+| status | text enum | 'pending' \| 'approved' \| 'in_progress' \| 'completed' \| 'blocked' \| 'pending_review' |
+| mergeBase | text nullable | git merge-base hash stored before phase merge, enables diff reconstruction for completed phases |
 | createdAt, updatedAt | integer/timestamp | |
 
 ### phase_dependencies
@@ -44,12 +45,13 @@ All adapters use nanoid() for IDs, auto-manage timestamps, and use Drizzle's `.r
 | parentTaskId | text nullable self-ref FK (cascade) | decomposition hierarchy |
 | name | text NOT NULL | |
 | description | text nullable | |
-| type | text enum | 'auto' \| 'checkpoint:human-verify' \| 'checkpoint:decision' \| 'checkpoint:human-action' |
+| type | text enum | 'auto' |
 | category | text enum | 'execute' \| 'research' \| 'discuss' \| 'plan' \| 'detail' \| 'refine' \| 'verify' \| 'merge' \| 'review' |
 | priority | text enum | 'low' \| 'medium' \| 'high' |
 | status | text enum | 'pending' \| 'in_progress' \| 'completed' \| 'blocked' |
 | order | integer | default 0 |
 | summary | text nullable | Agent result summary — propagated to dependent tasks as context |
+| retryCount | integer NOT NULL | default 0, incremented on agent crash auto-retry, reset on manual retry |
 | createdAt, updatedAt | integer/timestamp | |
 
 ### task_dependencies
@@ -70,6 +72,7 @@ All adapters use nanoid() for IDs, auto-manage timestamps, and use Drizzle's `.r
 | mode | text enum | 'execute' \| 'discuss' \| 'plan' \| 'detail' \| 'refine' |
 | pid | integer nullable | OS process ID |
 | exitCode | integer nullable | |
+| prompt | text nullable | Full assembled prompt passed to agent at spawn; persisted for durability after log cleanup |
 | outputFilePath | text nullable | |
 | result | text nullable | JSON |
 | pendingQuestions | text nullable | JSON |
@@ -194,6 +197,21 @@ Messages within a chat session.
 
 Index: `(chatSessionId)`.
 
+### errands
+
+Tracks errand work items linked to a project branch, optionally assigned to an agent.
+
+| Column | Type | Notes |
+|--------|------|-------|
+| id | text PK | caller-supplied |
+| description | text NOT NULL | human-readable description |
+| branch | text NOT NULL | working branch name |
+| baseBranch | text NOT NULL | default 'main' |
+| agentId | text FK → agents (set null) | assigned agent; null if unassigned |
+| projectId | text FK → projects (cascade) | owning project |
+| status | text enum | active, pending_review, conflict, merged, abandoned; default 'active' |
+| createdAt, updatedAt | integer/timestamp | |
+
 ### review_comments
 
 Inline review comments on phase diffs, persisted across page reloads.
@@ -214,7 +232,7 @@ Index: `(phaseId)`.
 
 ## Repository Interfaces
 
-13 repositories, each with standard CRUD plus domain-specific methods:
+14 repositories, each with standard CRUD plus domain-specific methods:
 
 | Repository | Key Methods |
 |-----------|-------------|
@@ -231,6 +249,7 @@ Index: `(phaseId)`.
 | ConversationRepository | create, findById, findPendingForAgent, answer |
 | ChatSessionRepository | createSession, findActiveSession, findActiveSessionByAgentId, updateSession, createMessage, findMessagesBySessionId |
 | ReviewCommentRepository | create, findByPhaseId, resolve, unresolve, delete |
+| ErrandRepository | create, findById, findAll (filter by projectId/status), update, delete |
 
 ## Migrations
 
@@ -242,4 +261,4 @@ Key rules:
 - See [database-migrations.md](database-migrations.md) for full workflow
 - Snapshots stale after 0008; migrations 0008+ are hand-written
 
-Current migrations: 0000 through 0030 (31 total).
+Current migrations: 0000 through 0035 (36 total).

docs/dispatch-events.md

@@ -11,7 +11,7 @@
 - **Adapter**: `TypedEventBus` using Node.js `EventEmitter`
 - All events implement `BaseEvent { type, timestamp, payload }`
 
-### Event Types (57)
+### Event Types (58)
 
 | Category | Events | Count |
 |----------|--------|-------|
@@ -27,7 +27,7 @@
 | **Preview** | `preview:building`, `preview:ready`, `preview:stopped`, `preview:failed` | 4 |
 | **Conversation** | `conversation:created`, `conversation:answered` | 2 | `conversation:created` triggers auto-resume of idle target agents via `resumeForConversation()` |
 | **Chat** | `chat:message_created`, `chat:session_closed` | 2 | Chat session lifecycle events |
-| **Initiative** | `initiative:pending_review`, `initiative:review_approved` | 2 | Initiative-level review gate events |
+| **Initiative** | `initiative:pending_review`, `initiative:review_approved`, `initiative:changes_requested` | 3 | Initiative-level review gate events |
 | **Project** | `project:synced`, `project:sync_failed` | 2 | Remote sync results from `ProjectSyncManager` |
 | **Log** | `log:entry` | 1 |
 
@@ -48,6 +48,7 @@ PhaseChangesRequestedEvent { phaseId, initiativeId, taskId, commentCount }
 AccountCredentialsRefreshedEvent { accountId, expiresAt, previousExpiresAt? }
 InitiativePendingReviewEvent   { initiativeId, branch }
 InitiativeReviewApprovedEvent  { initiativeId, branch, strategy: 'push_branch' | 'merge_and_push' }
+InitiativeChangesRequestedEvent { initiativeId, phaseId, taskId }
 ```
 
 ## Task Dispatch
@@ -64,10 +65,11 @@ InitiativeReviewApprovedEvent  { initiativeId, branch, strategy: 'push_branch' |
 2. **Dispatch** — `dispatchNext()` finds highest-priority task with all deps complete
 3. **Context gathering** — Before spawn, `dispatchNext()` gathers initiative context (initiative, phase, tasks, pages) and passes as `inputContext` to the agent. Agents receive `.cw/input/task.md`, `initiative.md`, `phase.md`, `context/tasks/`, `context/phases/`, and `pages/`.
 4. **Priority order**: high > medium > low, then oldest first (FIFO within priority)
-5. **Checkpoint skip** — Tasks with type starting with `checkpoint:` skip auto-dispatch
-6. **Planning skip** — Planning-category tasks (research, discuss, plan, detail, refine) skip auto-dispatch — they use the architect flow
+5. **Planning skip** — Planning-category tasks (research, discuss, plan, detail, refine) skip auto-dispatch — they use the architect flow
 7. **Summary propagation** — `completeTask()` reads the completing agent's `result.message` and stores it on the task's `summary` column. Dependent tasks see this summary in `context/tasks/<id>.md` frontmatter.
 8. **Spawn failure** — If `agentManager.spawn()` throws, the task is blocked via `blockTask()` with the error message. The dispatch cycle continues instead of crashing.
+9. **Retry blocked** — `retryBlockedTask(taskId)` resets a blocked task to pending and re-queues it. Exposed via tRPC `retryBlockedTask` mutation. The UI shows a Retry button in the task slide-over when status is `blocked`.
+10. **Branch validation** — Branch computation and `ensureBranch` errors are fatal for execution tasks (execute, verify, merge, review) but non-fatal for planning tasks. This prevents agents from spawning without proper branch isolation.
 
 ### DispatchManager Methods
 
@@ -78,6 +80,7 @@ InitiativeReviewApprovedEvent  { initiativeId, branch, strategy: 'push_branch' |
 | `getNextDispatchable()` | Get next task without dispatching |
 | `completeTask(taskId, agentId?)` | Complete task |
 | `blockTask(taskId, reason)` | Block task with reason |
+| `retryBlockedTask(taskId)` | Reset blocked task to pending and re-queue |
 | `getQueueState()` | Return queued, ready, blocked tasks |
 
 ## Phase Dispatch
@@ -110,8 +113,21 @@ InitiativeReviewApprovedEvent  { initiativeId, branch, strategy: 'push_branch' |
 | Event | Action |
 |-------|--------|
 | `phase:queued` | Dispatch ready phases → dispatch their tasks to idle agents |
-| `agent:stopped` | Re-dispatch queued tasks (freed agent slot) |
-| `task:completed` | Merge task branch, then dispatch next queued task |
+| `agent:stopped` | Auto-complete task (unless user_requested), re-dispatch queued tasks (freed agent slot) |
+| `agent:crashed` | Auto-retry crashed task up to `MAX_TASK_RETRIES` (3). Increments `retryCount`, resets status to `pending`, re-queues. Exceeding retries leaves task `in_progress` for manual intervention. |
+| `task:completed` | Merge task branch (if branch exists), check phase completion, dispatch next queued task |
+
+### Crash Recovery
+
+When an agent crashes (`agent:crashed` event), the orchestrator automatically retries the task:
+1. Finds the task associated with the crashed agent
+2. Checks `task.retryCount` against `MAX_TASK_RETRIES` (3)
+3. If under limit: increments `retryCount`, resets task to `pending`, re-queues for dispatch
+4. If over limit: logs warning, task stays `in_progress` for manual intervention
+
+On server restart, `recoverDispatchQueues()` also recovers stuck `in_progress` tasks whose agents are dead (status is not `running` or `waiting_for_input`). These are reset to `pending` and re-queued.
+
+Manual retry via `retryBlockedTask()` resets `retryCount` to 0, giving the task a fresh set of automatic retries.
 
 ### Coalesced Scheduling
 
@@ -119,6 +135,8 @@ Multiple rapid events (e.g. several `phase:queued` from `queueAllPhases`) are co
 
 ### Execution Mode Behavior
 
-- **YOLO**: phase completes → auto-merge → auto-dispatch next phase → auto-dispatch tasks
+- **YOLO**: phase completes → auto-merge (if branch exists, skipped otherwise) → auto-dispatch next phase → auto-dispatch tasks
 - **review_per_phase**: phase completes → set `pending_review` → STOP. User approves → `approveAndMergePhase()` → merge → dispatch next phase → dispatch tasks
-- **request-changes**: phase `pending_review` → user requests changes → creates revision task (category=`'review'`) → phase reset to `in_progress` → agent fixes → phase returns to `pending_review`
+- **No branch**: Phase completion check runs regardless of branch existence. Merge steps are skipped; status transitions still fire. `updateTaskStatus` tRPC routes completions through `dispatchManager.completeTask()` to emit `task:completed`.
+- **request-changes (phase)**: phase `pending_review` → user requests changes → creates revision task (category=`'review'`, dedup guard skips if active review exists) with threaded comments (`[comment:ID]` tags + reply threads) → phase reset to `in_progress` → agent reads comments, fixes code, writes `.cw/output/comment-responses.json` → OutputHandler creates reply comments and optionally resolves threads → phase returns to `pending_review`
+- **request-changes (initiative)**: initiative `pending_review` → user requests changes → creates/reuses "Finalization" phase → creates review task → initiative reset to `active` → agent fixes → initiative returns to `pending_review`

docs/frontend.md

@@ -44,6 +44,7 @@ Use `mapEntityStatus(rawStatus)` from `StatusDot.tsx` to convert raw entity stat
 |-------|-----------|---------|
 | `/` | `routes/index.tsx` | Dashboard / initiative list |
 | `/initiatives/$id` | `routes/initiatives/$initiativeId.tsx` | Initiative detail (tabbed) |
+| `/agents` | `routes/agents.tsx` | Agent list with Output / Details tab panel |
 | `/settings` | `routes/settings/index.tsx` | Settings page |
 
 ## Initiative Detail Tabs
@@ -54,7 +55,7 @@ The initiative detail page has three tabs managed via local state (not URL param
 2. **Execution Tab** — Pipeline visualization, phase management, task dispatch
 3. **Review Tab** — Pending proposals from agents
 
-## Component Inventory (73 components)
+## Component Inventory (74 components)
 
 ### Core Components (`src/components/`)
 | Component | Purpose |
@@ -66,6 +67,7 @@ The initiative detail page has three tabs managed via local state (not URL param
 | `StatusBadge` | Colored badge using status tokens |
 | `TaskRow` | Task list item with status, priority, category |
 | `QuestionForm` | Agent question form with options |
+| `AgentDetailsPanel` | Details tab for agent right-panel: metadata, input files, effective prompt |
 | `InboxDetailPanel` | Agent message detail + response form |
 | `ProjectPicker` | Checkbox list for project selection |
 | `RegisterProjectDialog` | Dialog to register new git project |
@@ -111,10 +113,12 @@ The initiative detail page has three tabs managed via local state (not URL param
 ### Review Components (`src/components/review/`)
 | Component | Purpose |
 |-----------|---------|
-| `ReviewTab` | Review tab container — orchestrates header, diff, sidebar, and preview |
+| `ReviewTab` | Review tab container — orchestrates header, diff, sidebar, and preview. Phase-level review has threaded inline comments (with reply support) + Request Changes; initiative-level review has Request Changes (summary prompt) + Push Branch / Merge & Push |
 | `ReviewHeader` | Consolidated toolbar: phase selector pills, branch info, stats, preview controls, approve/reject actions |
-| `ReviewSidebar` | VSCode-style icon strip (Files/Commits views) with file list, comment counts, and commit navigation |
-| `DiffViewer` | Unified diff renderer with inline comments |
+| `ReviewSidebar` | VSCode-style icon strip (Files/Commits views) with file list, root-only comment counts, and commit navigation |
+| `DiffViewer` | Unified diff renderer with threaded inline comments (root + reply threads) |
+| `CommentThread` | Renders root comment with resolve/reopen + nested reply threads (agent replies styled with primary border). Inline reply form |
+| `ConflictResolutionPanel` | Merge conflict detection + agent resolution in initiative review. Shows conflict files, spawns conflict agent, inline questions, re-check on completion |
 | `PreviewPanel` | Docker preview status: building/running/failed with start/stop (legacy, now integrated into ReviewHeader) |
 | `ProposalCard` | Individual proposal display |
 
@@ -126,6 +130,7 @@ shadcn/ui components: badge (6 status variants + xs size), button, card, dialog,
 | Hook | Purpose |
 |------|---------|
 | `useRefineAgent` | Manages refine agent lifecycle for initiative |
+| `useConflictAgent` | Manages conflict resolution agent lifecycle for initiative review |
 | `useDetailAgent` | Manages detail agent for phase planning |
 | `useAgentOutput` | Subscribes to live agent output stream |
 | `useChatSession` | Manages chat session for phase/task refinement |
@@ -195,4 +200,4 @@ Components: `ChatSlideOver`, `ChatBubble`, `ChatInput`, `ChangeSetInline` in `sr
 
 `listInitiatives` returns an `activity` field on each initiative, computed server-side from phase statuses via `deriveInitiativeActivity()` in `apps/server/trpc/routers/initiative-activity.ts`. This eliminates per-card N+1 `listPhases` queries.
 
-Activity states (priority order): active architect agents > `pending_review` > `executing` > `blocked` > `complete` > `ready` > `planning` > `idle` > `archived`. Each state maps to a `StatusVariant` + pulse animation in `InitiativeCard`'s `activityVisual()` function. Active architect agents (modes: discuss, plan, detail, refine) are checked first — mapping to `discussing`, `detailing`, `detailing`, `refining` states respectively — so auto-spawned agents surface activity even when no phases exist yet. `PhaseSidebarItem` also shows a spinner when a detail agent is active for its phase.
+Activity states (priority order): conflict agent > `archived` > active architect agents > `pending_review` > `executing` > `blocked` > `complete` > `ready` > `planning` > `idle`. Each state maps to a `StatusVariant` + pulse animation in `InitiativeCard`'s `activityVisual()` function. Active conflict agents (name starts with `conflict-`) are checked first — returning `resolving_conflict` (urgent variant, pulsing). Active architect agents (modes: discuss, plan, detail, refine) are checked next — mapping to `discussing`, `detailing`, `detailing`, `refining` states respectively — so auto-spawned agents surface activity even when no phases exist yet. `PhaseSidebarItem` also shows a spinner when a detail agent is active for its phase.

docs/git-process-logging.md

@@ -45,6 +45,9 @@ Worktrees stored in `.cw-worktrees/` subdirectory of the repo. Each agent gets a
 | `remoteBranchExists(repoPath, branch)` | Check remote tracking branches (`origin/<branch>`) |
 | `listCommits(repoPath, base, head)` | List commits head has that base doesn't (with stats) |
 | `diffCommit(repoPath, commitHash)` | Get unified diff for a single commit |
+| `getMergeBase(repoPath, branch1, branch2)` | Get common ancestor commit hash |
+| `pushBranch(repoPath, branch, remote?)` | Push branch to remote (default: 'origin') |
+| `checkMergeability(repoPath, source, target)` | Dry-run merge check via `git merge-tree --write-tree` (git 2.38+). Returns `{ mergeable, conflicts? }` with no side effects |
 
 `remoteBranchExists` is used by `registerProject` and `updateProject` to validate that a project's default branch actually exists in the cloned repository before saving.
 

docs/server-api.md

@@ -59,11 +59,15 @@ Each procedure uses `require*Repository(ctx)` helpers that throw `TRPCError(INTE
 | dismissAgent | mutation | Dismiss agent (set userDismissedAt) |
 | resumeAgent | mutation | Resume with answers |
 | listAgents | query | All agents |
-| getAgent | query | Single agent by name or ID |
+| getAgent | query | Single agent by name or ID; also returns `taskName`, `initiativeName`, `exitCode` |
 | getAgentResult | query | Execution result |
 | getAgentQuestions | query | Pending questions |
-| getAgentOutput | query | Full output from DB log chunks |
+| getAgentOutput | query | Timestamped log chunks from DB (`{ content, createdAt }[]`) |
+| getTaskAgent | query | Most recent agent assigned to a task (by taskId) |
+| getAgentInputFiles | query | Files written to agent's `.cw/input/` dir (text only, sorted, 500 KB cap) |
+| getAgentPrompt | query | Assembled prompt — reads from DB (`agents.prompt`) first; falls back to `.cw/agent-logs/<name>/PROMPT.md` for pre-persistence agents (1 MB cap) |
 | getActiveRefineAgent | query | Active refine agent for initiative |
+| getActiveConflictAgent | query | Active conflict resolution agent for initiative (name starts with `conflict-`) |
 | listWaitingAgents | query | Agents waiting for input |
 | onAgentOutput | subscription | Live raw JSONL output stream via EventBus |
 
@@ -95,6 +99,9 @@ Each procedure uses `require*Repository(ctx)` helpers that throw `TRPCError(INTE
 | getInitiativeReviewCommits | query | Commits on initiative branch not on default branch |
 | getInitiativeCommitDiff | query | Single commit diff for initiative review |
 | approveInitiativeReview | mutation | Approve initiative review: `{initiativeId, strategy: 'push_branch' \| 'merge_and_push'}` |
+| requestInitiativeChanges | mutation | Request changes on initiative: `{initiativeId, summary}` → creates review task in Finalization phase, resets initiative to active |
+| checkInitiativeMergeability | query | Dry-run merge check: `{initiativeId}` → `{mergeable, conflictFiles[], targetBranch}` |
+| spawnConflictResolutionAgent | mutation | Spawn agent to resolve merge conflicts: `{initiativeId, provider?}` → auto-dismisses stale conflict agents, creates merge task |
 
 ### Phases
 | Procedure | Type | Description |
@@ -115,11 +122,12 @@ Each procedure uses `require*Repository(ctx)` helpers that throw `TRPCError(INTE
 | getPhaseReviewCommits | query | List commits between initiative and phase branch |
 | getCommitDiff | query | Diff for a single commit (by hash) in a phase |
 | approvePhaseReview | mutation | Approve and merge phase branch |
-| requestPhaseChanges | mutation | Request changes: creates revision task from unresolved comments, resets phase to in_progress |
-| listReviewComments | query | List review comments by phaseId |
+| requestPhaseChanges | mutation | Request changes: creates revision task from unresolved threaded comments (with `[comment:ID]` tags and reply threads), resets phase to in_progress |
+| listReviewComments | query | List review comments by phaseId (flat list including replies, frontend groups by parentCommentId) |
 | createReviewComment | mutation | Create inline review comment on diff |
 | resolveReviewComment | mutation | Mark review comment as resolved |
 | unresolveReviewComment | mutation | Mark review comment as unresolved |
+| replyToReviewComment | mutation | Create a threaded reply to an existing review comment (copies parent's phaseId/filePath/lineNumber) |
 
 ### Phase Dispatch
 | Procedure | Type | Description |
@@ -205,13 +213,13 @@ Each procedure uses `require*Repository(ctx)` helpers that throw `TRPCError(INTE
 | Procedure | Type | Events |
 |-----------|------|--------|
 | onEvent | subscription | All event types |
-| onAgentUpdate | subscription | agent:* events (8 types) |
+| onAgentUpdate | subscription | agent:* events (7 types, excludes agent:output) |
 | onTaskUpdate | subscription | task:* + phase:* events (8 types) |
 | onPageUpdate | subscription | page:created/updated/deleted |
 | onPreviewUpdate | subscription | preview:building/ready/stopped/failed |
 | onConversationUpdate | subscription | conversation:created/answered |
 
-Subscriptions use `eventBusIterable()` — queue-based async generator, max 1000 events, 30s heartbeat.
+Subscriptions use `eventBusIterable()` — queue-based async generator, max 1000 events, 30s heartbeat. `agent:output` is excluded from all general subscriptions (it's high-frequency streaming data); use the dedicated `onAgentOutput` subscription instead.
 
 ## Coordination Module