
<!-- section: cli -->
<!-- Generated file. Do not edit. -->
<!-- eforge version: 0.7.12 -->
<!-- Commit: d616c57 -->
<!-- Source: packages/eforge/src/cli/index.ts -->

# eforge CLI Reference

Autonomous plan-build-review CLI for code generation.

**Usage:** `eforge [command] [options]`

## Commands

### `enqueue`

Normalize input and add it to the PRD queue


**Options:**

| Flag | Description |
|------|-------------|
| `--name <name>` | Override the inferred PRD title |
| `--verbose` | Stream agent output |
| `--no-plugins` | Disable plugin loading |
| `--profile <name>` | Override active profile for this enqueue + build |

### `build`

Compile + build + validate in one step

**Alias:** `run`


**Options:**

| Flag | Description |
|------|-------------|
| `--auto` | Run without approval gates |
| `--verbose` | Stream agent output |
| `--name <name>` | Plan set name (inferred from source if omitted) |
| `--queue` | Process all PRDs from the queue |
| `--max-concurrent-builds <n>` | Max parallel queue PRDs |
| `--dry-run` | Compile only, then show execution plan without building |
| `--foreground` | Run in-process instead of delegating to daemon |
| `--no-cleanup` | Keep plan files after successful build |
| `--no-monitor` | Disable web monitor |
| `--no-plugins` | Disable plugin loading |
| `--watch` | Watch mode: continuously poll the queue for new PRDs |
| `--poll-interval <ms>` | Poll interval in milliseconds for watch mode |

### `monitor`

Start or connect to the monitor dashboard


**Options:**

| Flag | Description |
|------|-------------|
| `--port <port>` | Preferred port |

### `status`

Check running builds


### `queue`

Manage PRD queue


#### `list`

Show PRDs in the queue


#### `run`

Process PRDs from the queue


**Options:**

| Flag | Description |
|------|-------------|
| `--all` | Process all pending PRDs |
| `--auto` | Run without approval gates |
| `--verbose` | Stream agent output |
| `--no-monitor` | Disable web monitor |
| `--no-plugins` | Disable plugin loading |
| `--max-concurrent-builds <n>` | Max parallel queue PRDs |
| `--watch` | Watch mode: continuously poll the queue for new PRDs |
| `--poll-interval <ms>` | Poll interval in milliseconds for watch mode |

#### `exec`

Build a single PRD directly (subprocess entry point for the queue scheduler)


**Options:**

| Flag | Description |
|------|-------------|
| `--auto` | Run without approval gates |
| `--verbose` | Stream agent output |
| `--no-monitor` | Disable web monitor |
| `--no-plugins` | Disable plugin loading |
| `--session-id <uuid>` | Session ID injected by parent scheduler (skips child session:start emission) |
| `--profile <name>` | Override active profile for this build |

### `config`

Manage eforge configuration


#### `validate`

Validate eforge/config.yaml configuration


#### `show`

Show resolved eforge configuration


### `debug-composer`

Run only the pipeline-composer stage under one or more backend profiles and dump the request payload each backend constructs (system prompt, tools, model, thinking) for side-by-side diffing. Use --backend <name> to select profiles; repeat to compare multiple.


**Options:**

| Flag | Description |
|------|-------------|
| `--backend <name>` | Backend profile to test (repeatable). Defaults to the currently-active profile. |
| `--out <dir>` | Output directory for captured payloads |
| `--verbose` | Stream composer agent messages to stdout |

### `daemon`

Manage persistent daemon server


#### `start`

Start the persistent daemon server


**Options:**

| Flag | Description |
|------|-------------|
| `--port <port>` | Preferred port |

#### `stop`

Stop the persistent daemon server


**Options:**

| Flag | Description |
|------|-------------|
| `--force` | Skip active-build safety check |

#### `status`

Show daemon status


#### `kill`

Force-kill the daemon (SIGKILL)


### `recover`

Analyse a failed build and write recovery sidecar files


**Options:**

| Flag | Description |
|------|-------------|
| `--cwd <cwd>` | Working directory override |
| `--verbose` | Stream agent output |
| `--no-monitor` | Disable web monitor |

### `apply-recovery`

Apply the recovery verdict for a failed build plan (requeue, enqueue successor, or abandon)


**Options:**

| Flag | Description |
|------|-------------|
| `--cwd <cwd>` | Working directory override |
| `--no-monitor` | Disable web monitor |

### `playbook`

Manage playbooks


#### `list`

List all available playbooks with source and shadow chain


#### `new`

Scaffold a new playbook (non-interactive, for scripts)


**Options:**

| Flag | Description |
|------|-------------|
| `--scope <scope>` | Playbook scope: user \| project-team \| project-local |
| `--name <name>` | Playbook name (kebab-case) |
| `--description <description>` | Short description of the playbook |
| `--from <file>` | Read body content from this file (used as the Goal section) |

#### `edit`

Open a playbook in $EDITOR, validate, and save to the same tier


#### `run`

Enqueue a playbook as a PRD


**Options:**

| Flag | Description |
|------|-------------|
| `--after <queue-id>` | Queue ID that this PRD should run after (piggyback) |

#### `promote`

Promote a playbook from project-local to project-team (stages with git add)


#### `demote`

Demote a playbook from project-team to project-local (.eforge/playbooks/)


### `play`

Shortcut for `eforge playbook run <name>`


**Options:**

| Flag | Description |
|------|-------------|
| `--after <queue-id>` | Queue ID that this PRD should run after (piggyback) |

### `mcp-proxy`

Run the MCP stdio proxy server (used by Claude Code plugin)


## Global options

**Options:**

| Flag | Description |
|------|-------------|
| `-V, --version` | output the version number |

<!-- end-section: cli -->

<!-- section: api -->
<!-- Generated file. Do not edit. -->
<!-- eforge version: 0.7.12 -->
<!-- Commit: d616c57 -->
<!-- Source: packages/client/src/routes.ts -->

# eforge Daemon HTTP API Reference

The eforge daemon exposes an HTTP API at `http://localhost:{port}/api/...`.
Clients should import route constants from `@eforge-build/client` (`API_ROUTES`) rather
than embedding literal path strings.

## Routes

Total routes: 48

| Route key | Path pattern |
|-----------|-------------|
| `applyRecovery` | `/api/recover/apply` |
| `autoBuildGet` | `/api/auto-build` |
| `autoBuildSet` | `/api/auto-build` |
| `cancel` | `/api/cancel/:sessionId` |
| `configShow` | `/api/config/show` |
| `configValidate` | `/api/config/validate` |
| `daemonEvents` | `/api/daemon-events` |
| `daemonStop` | `/api/daemon/stop` |
| `diff` | `/api/diff/:sessionId/:planId` |
| `enqueue` | `/api/enqueue` |
| `events` | `/api/events/:runId` |
| `health` | `/api/health` |
| `keepAlive` | `/api/keep-alive` |
| `modelList` | `/api/models/list` |
| `modelProviders` | `/api/models/providers` |
| `plans` | `/api/plans/:runId` |
| `playbookCopy` | `/api/playbook/copy` |
| `playbookDemote` | `/api/playbook/demote` |
| `playbookEnqueue` | `/api/playbook/enqueue` |
| `playbookList` | `/api/playbook/list` |
| `playbookPromote` | `/api/playbook/promote` |
| `playbookSave` | `/api/playbook/save` |
| `playbookShow` | `/api/playbook/show` |
| `playbookValidate` | `/api/playbook/validate` |
| `profileCreate` | `/api/profile/create` |
| `profileDelete` | `/api/profile/:name` |
| `profileList` | `/api/profile/list` |
| `profileShow` | `/api/profile/show` |
| `profileUse` | `/api/profile/use` |
| `projectContext` | `/api/project-context` |
| `queue` | `/api/queue` |
| `readRecoverySidecar` | `/api/recovery/sidecar` |
| `recover` | `/api/recover` |
| `runs` | `/api/runs` |
| `runState` | `/api/run-state/:id` |
| `runSummary` | `/api/run-summary/:id` |
| `schedulerKick` | `/api/scheduler/kick` |
| `sessionMetadata` | `/api/session-metadata` |
| `sessionPlanCreate` | `/api/session-plan/create` |
| `sessionPlanList` | `/api/session-plan/list` |
| `sessionPlanMigrateLegacy` | `/api/session-plan/migrate-legacy` |
| `sessionPlanReadiness` | `/api/session-plan/readiness` |
| `sessionPlanSelectDimensions` | `/api/session-plan/select-dimensions` |
| `sessionPlanSetSection` | `/api/session-plan/set-section` |
| `sessionPlanSetStatus` | `/api/session-plan/set-status` |
| `sessionPlanShow` | `/api/session-plan/show` |
| `sessionPlanSkipDimension` | `/api/session-plan/skip-dimension` |
| `version` | `/api/version` |

## SSE Streams

- `GET /api/daemon-events` — daemon-wide event stream with `stream:hello` snapshot on connect.
- `GET /api/events/:runId` — session-specific event stream with `stream:hello` snapshot on connect.

Use `buildPath(pattern, params)` from `@eforge-build/client` to resolve `:param` placeholders.

<!-- end-section: api -->

<!-- section: events -->
<!-- Generated file. Do not edit. -->
<!-- eforge version: 0.7.12 -->
<!-- Commit: d616c57 -->
<!-- Source: packages/client/src/events.schemas.ts -->

# eforge Event Protocol Reference

All events emitted on the eforge SSE stream conform to the `EforgeEvent` discriminated
union defined in `packages/client/src/events.schemas.ts`.

Each event carries an optional envelope (`sessionId`, `runId`, `timestamp`) intersected
with one of the variant objects below. The `type` field discriminates the variant.

## Event Variants

Total variants: 150

| Event type | Additional fields |
|------------|-------------------|
| `session:start` | `sessionId` |
| `session:end` | `result`, `sessionId` |
| `session:profile` | `config`, `profileName`, `scope`, `source` |
| `phase:start` | `command`, `planSet`, `runId` |
| `phase:end` | `result`, `runId` |
| `config:warning` | `details`, `message`, `source` |
| `planning:warning` | `details`, `message`, `planId`, `source` |
| `planning:module:build-config:invalid` | `errors`, `moduleId`, `reason` |
| `planning:start` | `label`, `source` |
| `planning:skip` | `reason` |
| `planning:submission` | `hasMigrations`, `planCount`, `totalBodySize` |
| `planning:error` | `reason` |
| `planning:clarification` | `questions` |
| `planning:clarification:answer` | `answers` |
| `planning:progress` | `message` |
| `planning:continuation` | `attempt`, `maxContinuations`, `reason` |
| `planning:pipeline` | `compile`, `defaultBuild`, `defaultReview`, `rationale`, `scope` |
| `planning:complete` | `planConfigs`, `plans` |
| `planning:review:start` | - |
| `planning:review:complete` | `issues` |
| `planning:evaluate:start` | - |
| `planning:evaluate:continuation` | `attempt`, `maxContinuations` |
| `planning:evaluate:complete` | `accepted`, `rejected`, `verdicts` |
| `planning:architecture:review:start` | - |
| `planning:architecture:review:complete` | `issues` |
| `planning:architecture:evaluate:start` | - |
| `planning:architecture:evaluate:continuation` | `attempt`, `maxContinuations` |
| `planning:architecture:evaluate:complete` | `accepted`, `rejected`, `verdicts` |
| `planning:cohesion:start` | - |
| `planning:cohesion:complete` | `issues` |
| `planning:cohesion:evaluate:start` | - |
| `planning:cohesion:evaluate:continuation` | `attempt`, `maxContinuations` |
| `planning:cohesion:evaluate:complete` | `accepted`, `rejected`, `verdicts` |
| `plan:build:start` | `planId` |
| `plan:build:implement:start` | `planId` |
| `plan:build:implement:progress` | `message`, `planId` |
| `plan:build:implement:continuation` | `attempt`, `maxContinuations`, `planId`, `shardId` |
| `plan:build:implement:complete` | `planId` |
| `plan:build:files_changed` | `baseBranch`, `diffs`, `files`, `planId` |
| `plan:build:review:start` | `planId` |
| `plan:build:review:complete` | `issues`, `planId` |
| `plan:build:review:parallel:start` | `perspectives`, `planId` |
| `plan:build:review:parallel:perspective:start` | `perspective`, `planId` |
| `plan:build:review:parallel:perspective:complete` | `issues`, `perspective`, `planId` |
| `plan:build:review:parallel:perspective:error` | `error`, `perspective`, `planId` |
| `plan:build:review:fix:start` | `issueCount`, `planId` |
| `plan:build:review:fix:complete` | `planId` |
| `plan:build:evaluate:start` | `planId` |
| `plan:build:evaluate:continuation` | `attempt`, `maxContinuations`, `planId` |
| `plan:build:evaluate:complete` | `accepted`, `planId`, `rejected`, `verdicts` |
| `plan:build:doc-author:start` | `planId` |
| `plan:build:doc-author:complete` | `docsAuthored`, `planId` |
| `plan:build:doc-sync:start` | `planId` |
| `plan:build:doc-sync:complete` | `docsSynced`, `planId` |
| `plan:build:test:write:start` | `planId` |
| `plan:build:test:write:complete` | `planId`, `testsWritten` |
| `plan:build:test:start` | `planId` |
| `plan:build:test:complete` | `failed`, `passed`, `planId`, `productionIssues`, `testBugsFixed` |
| `plan:build:complete` | `planId` |
| `plan:build:failed` | `error`, `planId`, `terminalSubtype` |
| `plan:build:progress` | `message`, `planId` |
| `plan:status:change` | `planId`, `status` |
| `plan:error:set` | `error`, `planId` |
| `plan:error:clear` | `planId` |
| `schedule:start` | `planIds` |
| `plan:schedule:ready` | `planId`, `reason` |
| `plan:merge:start` | `planId` |
| `plan:merge:complete` | `commitSha`, `planId` |
| `plan:merge:resolve:start` | `planId` |
| `plan:merge:resolve:complete` | `planId`, `resolved` |
| `merge:finalize:start` | `baseBranch`, `featureBranch` |
| `merge:finalize:complete` | `baseBranch`, `commitSha`, `featureBranch` |
| `merge:finalize:skipped` | `baseBranch`, `featureBranch`, `reason` |
| `merge:worktree:set` | `path` |
| `merge:worktree:clear` | - |
| `expedition:architecture:complete` | `modules` |
| `expedition:wave:start` | `moduleIds`, `wave` |
| `expedition:wave:complete` | `wave` |
| `expedition:module:start` | `moduleId` |
| `expedition:module:complete` | `moduleId` |
| `expedition:compile:start` | - |
| `expedition:compile:complete` | `plans` |
| `agent:start` | `agent`, `agentId`, `effort`, `effortClamped`, `effortOriginal`, `effortSource`, `harness`, `harnessSource`, `model`, `perspective`, `planId`, `projectMcpSelection`, `projectMcpServerNames`, `thinking`, `thinkingCoerced`, `thinkingOriginal`, `thinkingSource`, `tier`, `tierSource`, `toolbelt`, `toolbeltSource` |
| `agent:warning` | `agent`, `agentId`, `code`, `message`, `planId` |
| `agent:stop` | `agent`, `agentId`, `error`, `planId` |
| `agent:usage` | `agent`, `agentId`, `costUsd`, `final`, `numTurns`, `planId`, `usage` |
| `agent:message` | `agent`, `agentId`, `content`, `planId` |
| `agent:tool_use` | `agent`, `agentId`, `input`, `planId`, `tool`, `toolUseId` |
| `agent:tool_result` | `agent`, `agentId`, `output`, `planId`, `tool`, `toolUseId` |
| `agent:result` | `agent`, `planId`, `result` |
| `agent:retry` | `agent`, `attempt`, `label`, `maxAttempts`, `planId`, `shardId`, `subtype` |
| `validation:start` | `commands` |
| `validation:command:start` | `command` |
| `validation:command:complete` | `command`, `exitCode`, `output` |
| `validation:command:timeout` | `command`, `pid`, `timeoutMs` |
| `validation:complete` | `passed` |
| `validation:fix:start` | `attempt`, `maxAttempts` |
| `validation:fix:complete` | `attempt` |
| `prd_validation:start` | - |
| `prd_validation:complete` | `completionPercent`, `gaps`, `passed` |
| `gap_close:start` | `completionPercent`, `gapCount` |
| `gap_close:plan_ready` | `gaps`, `planBody` |
| `gap_close:complete` | `passed` |
| `reconciliation:start` | - |
| `reconciliation:complete` | `report` |
| `cleanup:start` | `planSet` |
| `cleanup:complete` | `planSet` |
| `approval:needed` | `action`, `details`, `planId` |
| `approval:response` | `approved` |
| `enqueue:start` | `source` |
| `enqueue:complete` | `filePath`, `id`, `planSet`, `title` |
| `enqueue:failed` | `error` |
| `enqueue:commit-failed` | `error` |
| `recovery:start` | `prdId`, `setName` |
| `recovery:summary` | `prdId`, `summary` |
| `recovery:complete` | `prdId`, `sidecarJsonPath`, `sidecarMdPath`, `verdict` |
| `recovery:error` | `error`, `prdId`, `rawOutput` |
| `recovery:apply:start` | `prdId` |
| `recovery:apply:complete` | `noAction`, `prdId`, `successorPrdId`, `verdict` |
| `recovery:apply:error` | `message`, `prdId` |
| `daemon:run:upsert` | `run` |
| `daemon:auto-build:paused` | `reason` |
| `daemon:lifecycle:starting` | `mode`, `pid`, `port`, `version` |
| `daemon:lifecycle:ready` | `mode`, `pid`, `port`, `recoveryDurationMs`, `version` |
| `daemon:lifecycle:shutdown:start` | `reason`, `signal` |
| `daemon:lifecycle:shutdown:complete` | `durationMs` |
| `daemon:heartbeat` | `autoBuild`, `queueDepth`, `runningBuilds`, `subscribers`, `uptime` |
| `daemon:scheduler:dequeued` | `capacityRemaining`, `prdId`, `queueDepth` |
| `daemon:scheduler:capacity-blocked` | `limit`, `queueDepth`, `runningCount` |
| `daemon:scheduler:dependency-blocked` | `blockedBy`, `prdId` |
| `daemon:auto-build:enabled` | - |
| `daemon:auto-build:resumed` | - |
| `daemon:auto-build:triggered` | `prdsEnqueued`, `trigger` |
| `daemon:recovery:start` | - |
| `daemon:recovery:run-marked-failed` | `planSet`, `reason`, `runId` |
| `daemon:recovery:lock-removed` | `path`, `pid` |
| `daemon:recovery:complete` | `durationMs`, `locksRemoved`, `runsFailed` |
| `daemon:orphan:reaped` | `pid`, `planSet`, `runId`, `sessionId` |
| `daemon:warning` | `details`, `message`, `source` |
| `daemon:error` | `message`, `source`, `stack` |
| `queue:start` | `dir`, `prdCount` |
| `queue:prd:start` | `prdId`, `title` |
| `queue:prd:discovered` | `prdId`, `title` |
| `queue:prd:stale` | `justification`, `prdId`, `revision`, `title`, `verdict` |
| `queue:prd:skip` | `prdId`, `reason` |
| `queue:prd:commit-failed` | `error`, `prdId`, `title` |
| `queue:prd:complete` | `prdId`, `status` |
| `queue:complete` | `processed`, `skipped` |
| `plan:build:decision` | `decision`, `planId` |
| `planning:decision` | `decision`, `planId` |

## JSON Schema

The complete machine-readable schema is at [`/schemas/events.schema.json`](/schemas/events.schema.json).
Use `safeParseEforgeEvent(value)` from `@eforge-build/client` to validate at runtime.

<!-- end-section: events -->

<!-- section: config -->
<!-- Generated file. Do not edit. -->
<!-- eforge version: 0.7.12 -->
<!-- Commit: d616c57 -->
<!-- Source: packages/engine/src/config.ts -->

# eforge Configuration Reference

eforge merges configuration from three tiers (highest precedence first):

1. `.eforge/config.yaml` — project-local, gitignored, developer-personal
2. `eforge/config.yaml` — project-level, committed
3. `~/.config/eforge/config.yaml` — user-global

## Top-level fields

| Field | Description |
|-------|-------------|
| `agents` |  |
| `build` |  |
| `daemon` |  |
| `hooks` |  |
| `langfuse` |  |
| `maxConcurrentBuilds` |  |
| `monitor` |  |
| `plan` |  |
| `plugins` |  |
| `prdQueue` |  |
| `tools` |  |

## JSON Schema

The complete machine-readable schema is at [`/schemas/config.schema.json`](/schemas/config.schema.json).

<!-- end-section: config -->

<!-- section: tools -->
<!-- Generated file. Do not edit. -->
<!-- eforge version: 0.7.12 -->
<!-- Commit: d616c57 -->
<!-- Source: packages/eforge/src/cli/mcp-proxy.ts, packages/pi-eforge/extensions/eforge/index.ts, eforge-plugin/skills/, packages/pi-eforge/skills/ -->

# eforge MCP Tools and Skills Reference

eforge exposes its capabilities through two integration surfaces:
- **MCP tools** for the Claude Code plugin (`eforge-plugin/`)
- **Native Pi commands and tools** for the Pi extension (`packages/pi-eforge/`)

Both surfaces are kept in parity per `AGENTS.md`.

## MCP tools (Claude Code)

Total tools: 15

| Tool name | Description |
|-----------|-------------|
| `eforge_build` | Enqueue a PRD source for the eforge daemon to build. Returns a sessionId and autoBuild status. |
| `eforge_follow` | Follow a running eforge session: streams phase/files-changed/issue updates as progress notifications and returns the final session summary. Use after eforge_build to surface live build status in the conversation. |
| `eforge_auto_build` | Get or set the daemon auto-build state. When enabled, the daemon automatically builds PRDs as they are enqueued. |
| `eforge_status` | Get the current run status including plan progress, session state, event summary, and the daemon vs CLI version. |
| `eforge_queue_list` | List all PRDs currently in the eforge queue with their metadata. |
| `eforge_config` | Show resolved eforge configuration or validate eforge/config.yaml. Config merges three tiers: user (~/.config/eforge/config.yaml), project (eforge/config.yaml), and project-local (.eforge/config.yaml, gitignored). Pass verbose: true with action "show" to see per-tier file presence. |
| `eforge_profile` | Manage named profiles in eforge/profiles/ (project), .eforge/profiles/ (local, gitignored), or ~/.config/eforge/profiles/ (user). Actions: "list" enumerates profiles and reports which is active; "show" returns the resolved active profile; "use" writes the active-profile marker to switch profiles; "create" writes a new profile (pass `agents.tiers` with self-contained tier recipes; optionally pass `metadata` with `description`, `whenToUse`, and `tags` — descriptive only, does not affect runtime behavior); "delete" removes a profile (refuses when active unless force: true). |
| `eforge_models` | List providers or models available for a given harness. Actions: "providers" returns provider names (claude-sdk is implicit / returns []); "list" returns models, optionally filtered to a single provider, newest-first. |
| `eforge_daemon` | Manage the eforge daemon lifecycle: start, stop, or restart the daemon. |
| `eforge_init` | Initialize eforge in a project. The skill is responsible for picking harness/model/effort per tier interactively; the tool is a pure persister. Pass `profile.tiers` with one self-contained recipe per tier (planning/implementation/review/evaluation). Each tier carries its own harness + model + effort. |
| `eforge_recover` | Trigger failure recovery analysis for a failed build plan. Spawns the recovery agent as a background subprocess and returns its sessionId and pid. |
| `eforge_read_recovery_sidecar` | Read the recovery analysis sidecar files for a failed build plan. Returns both the markdown summary and the structured JSON verdict produced by the recovery agent. |
| `eforge_apply_recovery` | Apply the recovery verdict for a failed build plan: requeue (retry), enqueue successor (split), or archive (abandon). |
| `eforge_playbook` | Manage playbooks in eforge. Actions: "list" returns all playbooks with source and shadow chain; "show" returns a single playbook's frontmatter and body; "save" validates and writes a playbook to the target tier; "enqueue" loads a playbook and enqueues it as a PRD, optionally chained after another queue entry; "promote" moves a playbook from project-local (.eforge/playbooks/) to project-team (eforge/playbooks/); "demote" reverses a promote; "validate" checks a raw Markdown playbook string without writing. |
| `eforge_session_plan` | Manage session plans in eforge. Actions: "list-active" returns all active (planning/ready) session plans; "show" returns a single session plan's data and readiness detail; "create" creates a new session plan file; "set-section" writes a dimension section to the session file; "skip-dimension" records a skipped dimension with a reason; "set-status" updates the session plan status (e.g. to "ready" or "abandoned"); "select-dimensions" sets planning type and depth and populates the required/optional dimension lists from the work-type playbook; "readiness" checks whether all required dimensions are covered; "migrate-legacy" converts a legacy boolean-dimensions session file to the current shape. Pass open: true on "create" or "show" to best-effort open the session plan file in the default application. |

## Native tools (Pi extension)

Total tools: 16

| Tool name | Description |
|-----------|-------------|
| `eforge_build` | Enqueue a PRD source for the eforge daemon to build. Returns a sessionId and autoBuild status. |
| `eforge_follow` | Follow a running eforge session: streams phase/files-changed/issue updates as tool progress messages and returns the final session summary. Use after eforge_build to surface live build status in the conversation. |
| `eforge_status` | Get the current run status including plan progress, session state, event summary, and the daemon vs Pi-extension version. |
| `eforge_queue_list` | List all PRDs currently in the eforge queue with their metadata. |
| `eforge_config` | Show resolved eforge configuration or validate eforge/config.yaml. |
| `eforge_profile` | Manage named profiles in eforge/profiles/. Actions: "list" enumerates profiles and reports which is active; "show" returns the resolved active profile with harness; "use" writes eforge/.active-profile to switch profiles; "create" writes a new eforge/profiles/<name>.yaml (optionally pass `metadata` with `description`, `whenToUse`, and `tags` — descriptive only, does not affect runtime behavior); "delete" removes a profile (refuses when active unless force: true). |
| `eforge_models` | List providers or models available for a given harness. Actions: "providers" returns provider names (claude-sdk is implicit / returns []); "list" returns models, optionally filtered to a single provider, newest-first. |
| `eforge_daemon` | Manage the eforge daemon lifecycle: start, stop, or restart the daemon. |
| `eforge_auto_build` | Get or set the daemon auto-build state. When enabled, the daemon automatically builds PRDs as they are enqueued. |
| `eforge_init` | Initialize eforge in a project. The skill is responsible for picking provider/model interactively; the tool is a pure persister. Pass `profile` with the assembled multi-runtime spec (every runtime must use harness: 'pi'). With migrate: true, extracts legacy harness config from a pre-overhaul config.yaml. |
| `eforge_confirm_build` | Present an interactive TUI overlay for the user to confirm, edit, or cancel a build source before enqueuing. Returns the user's choice. |
| `eforge_recover` | Trigger failure recovery analysis for a failed build plan. Spawns the recovery agent as a background subprocess and returns its sessionId and pid. |
| `eforge_read_recovery_sidecar` | Read the recovery analysis sidecar files for a failed build plan. Returns both the markdown summary and the structured JSON verdict produced by the recovery agent. |
| `eforge_apply_recovery` |  |
| `eforge_playbook` | Manage playbooks in eforge. Actions: "list" returns all playbooks with source and shadow chain; "show" returns a single playbook's frontmatter and body; "save" validates and writes a playbook to the target tier; "enqueue" loads a playbook and enqueues it as a PRD, optionally chained after another queue entry; "promote" moves a playbook from project-local (.eforge/playbooks/) to project-team (eforge/playbooks/); "demote" reverses a promote; "validate" checks a raw Markdown playbook string without writing. |
| `eforge_session_plan` | Manage session plans in eforge. Actions: "list-active" returns all active (planning/ready) session plans; "show" returns a single session plan's data and readiness detail; "create" creates a new session plan file; "set-section" writes a dimension section to the session file; "skip-dimension" records a skipped dimension with a reason; "set-status" updates the session plan status (e.g. to "ready" or "abandoned"); "select-dimensions" sets planning type and depth and populates the required/optional dimension lists from the work-type playbook; "readiness" checks whether all required dimensions are covered; "migrate-legacy" converts a legacy boolean-dimensions session file to the current shape. Pass open: true on "create" or "show" to best-effort open the session plan file in the default application. |

## Skill surfaces

Slash-command skills for Claude Code (plugin) and Pi are kept in parity.
Source of truth: `scripts/check-skill-parity.mjs`.

| Skill (Claude Code `/eforge:<name>`) | Skill (Pi `eforge:<name>`) | Description |
|--------------------------------------|---------------------------|-------------|
| `profile` | `eforge-profile` | List, inspect, and switch agent runtime profiles |
| `profile-new` | `eforge-profile-new` | Create a new agent runtime profile in eforge/profiles/ |
| `build` | `eforge-build` | Enqueue a source for the eforge daemon to build via MCP tool |
| `config` | `eforge-config` | Initialize or edit eforge/config.yaml team-wide settings, with validation via MCP tool |
| `init` | `eforge-init` | Initialize eforge in the current project with an interactive setup form |
| `plan` | `eforge-plan` | Start or resume a structured planning conversation for changes to be built by eforge. Classifies work type and depth, selects relevant dimensions from a per-type playbook, captures acceptance criteria, and produces a session plan that /eforge:build enqueues. |
| `restart` | `eforge-restart` | Safely restart the eforge daemon, checking for active builds first |
| `status` | `eforge-status` | Check eforge run status and queue state via MCP tools |
| `update` | `eforge-update` | Check for eforge updates and guide through updating the CLI package, daemon, and plugin |
| `playbook` | `eforge-playbook` | Create, edit, run, list, and promote eforge playbooks — reusable recurring-workflow templates |
| `recover` | `eforge-recover` | Inspect the recovery verdict for a failed PRD and apply the recommended action (retry, split, or abandon) |

<!-- end-section: tools -->