deepwork-status
View deepwork session status — phase, team, bar verdicts, proposals, guardrails
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Menú
View deepwork session status — phase, team, bar verdicts, proposals, guardrails
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
Lightweight single-gate amendment for execute mode — spawns MICRO-TEAM (CRITIC + 1 specialist) for re-verdict without full deepwork re-run
View execute-mode session status — phase, plan_hash, drift, change_log, test results, 3-dimension verdict table, rollback log, discoveries
Run a deepwork session: research/design convergence (default) or plan execution (--mode execute). Design mode spawns a 5-archetype oppositional team and delivers an approved plan. Execute mode drives faithful implementation of an approved plan via role-asymmetric agents.
End an active deepwork session — graceful per-teammate shutdown, archives state, and restores settings. CLI auto-cleans team dirs at session end. Works for both mid-flight abort and post-HALT cleanup.
Karpathy-style recap of deepwork history — reads DEEPWORK_WIKI.md Overview and the last 3 log entries, outputs 30-50 plain-text words
Add, remove, or list written-bar criteria for the active deepwork session
| name | deepwork-status |
| description | View deepwork session status — phase, team, bar verdicts, proposals, guardrails |
| allowed-tools | ["Read(${CLAUDE_PROJECT_DIR:-$(pwd -P)}/.claude/deepwork/**)","Glob","TaskList"] |
Display the current status of all active deepwork sessions.
Glob: ${CLAUDE_PROJECT_DIR:-$(pwd -P)}/.claude/deepwork/*/state.json
If no files are found, report that no deepwork session is currently active and stop.
| Instance ID | Goal | Phase | Team |
|---|---|---|---|
<id> | <goal> | <phase> | <team_name> |
For the current session's instance (match state.json.session_id against $CLAUDE_CODE_SESSION_ID if available, or show all if the env var is unavailable), display detailed status:
Core fields
Goal
Phase (scope | explore | synthesize | critique | refine | deliver | done | refining)
Team: team_name
Instance: instance_id
Started: started_at
Last updated: last_updated
Safe mode: safe_mode
Source of truth (source_of_truth[]): list each path; or "(none specified)"
Anchors (anchors[]): list each file:line; or "(none specified)"
Written bar (bar[]): table of:
| ID | Criterion | Verdict | Categorical ban | Evidence required |
|---|
Empirical unknowns (empirical_unknowns[]): table of:
| ID | Description | Artifact | Owner | Result |
|---|
Role definitions (role_definitions[]): table of:
| Name | Archetype | Model | Output artifact |
|---|
Guardrails (guardrails[]): list:
- <rule> [source: <source>, <timestamp>]
Banners (banners[]): if non-empty, display:
Banners: N artifacts flagged
- <artifact_path>: <reason> (<banner_type>, <added_at>)
...
If empty or absent, omit this section entirely.
Proposal versions (list proposals/*.md files with their version: and delta_from_prior: front-matter if available).
Cross-check state (runs after step 3, before TaskList):
A. Glob the active instance dir: ${CLAUDE_PROJECT_DIR:-$(pwd -P)}/.claude/deepwork/<id>/*.md (NOT archived — live session).
B. For each file, extract the frontmatter (first --- block).
C. From each parsed frontmatter, collect:
- artifact_type
- bar_id (or bar_ids)
- cross_check_for (findings artifacts — points to the bar_id being cross-checked)
- verdict (critique artifacts — HOLDING|APPROVED)
- result (empirical_results — confirmed|refuted|inconclusive)
D. Render a "Cross-check state" column in the bar verdict table:
| Gate | Verdict | Cross-check state | Evidence |
|---|---|---|---|
| G6 | PASS | 2/2 confirmed | findings.hook-auditor.md (primary), mechanism.enforcement-designer.md (secondary) |
| G3 | PASS | empirical confirmed | empirical_results.E2.md (result: confirmed) |
| E. Backwards compatible: artifacts without frontmatter contribute no cross-check state; | |||
column shows —. | |||
| F. Also display: |
critique.v*.md (highest version), read verdict field.
Display: "CRITIC verdict: <verdict> (critique.v<version>.md)"empirical_results.*.md, read empirical_id + result.
Display table:
| Unknown | Result |
| <empirical_id> | <result> |Hooks health check: Read state.json.hooks_inject_status. Display under Hooks section:
hooks_inject_status is present: Injected: <block_count> blocks at <timestamp> — no action needed.hooks_inject_status is absent or null: display a warning: WARNING: hooks_inject_status missing — hooks may not have been injected. Run /deepwork-teardown and restart the session if enforcement hooks are required.hooks_inject_status.timestamp against started_at: if the inject timestamp predates started_at by more than 60 seconds, display: WARNING: hooks_inject_status is older than session start — may reflect a prior session's injection.Call TaskList to get live task status. Display tasks grouped by status (in_progress, pending, completed) with their owner and metadata.bar_id where available.
Read the last 50 lines of ${CLAUDE_PROJECT_DIR:-$(pwd -P)}/.claude/deepwork/<id>/log.md and display as Recent Activity.
If state.json.hook_warnings[] is non-empty, display under Hook Warnings (these are incident signals that triggered guardrail auto-append).
If state.json.user_feedback is set, display it under User Feedback (from a prior ExitPlanMode rejection).
Format all of this as a clean, readable dashboard. Use markdown tables; keep the overall output scannable.