Agent Skills: Deepwork Execute Status

Deepwork Execute Status

Display the current status of the active execute-mode deepwork session.

1. Use Glob to find active instances:

Glob: ${CLAUDE_PROJECT_DIR:-$(pwd -P)}/.claude/deepwork/*/state.json

If no files are found, report "no deepwork session is currently active" and stop. If multiple are found, identify the execute-mode instance by state.json.execute.phase being non-null and not "halt". If none has execute mode active, report "no execute-mode session is currently active" and stop.

2. Read the identified state.json. Display a header summary:

| Instance ID | Goal | Execute Phase | Team | |---|---|---|---| | <id> | <goal> | <execute.phase> | <team_name> |

3. Display Execute Core State:

   • Goal: goal
   • Execute Phase: execute.phase
   • Plan Ref: execute.plan_ref
   • Plan Hash: execute.plan_hash
   • Plan Drift: execute.plan_drift_detected — if true, display DRIFT DETECTED — resolve via /deepwork-execute-amend before proceeding
   • Started: started_at
   • Last Updated: last_updated

4. Display Authorization Flags (flat fields under execute):

   | Flag | Value | |---|---| | execute.authorized_push | <value> | | execute.authorized_force_push | <value> | | execute.authorized_prod_deploy | <value> | | execute.authorized_local_destructive | <value> | | execute.secret_scan_waived | <value> |

   If all flags are absent, display "(not yet set — SETUP incomplete)".

5. Display Written Bar (bar[]):

   | ID | Criterion | Verdict | Categorical Ban | Evidence Required | |---|---|---|---|---|

   Or "(not yet populated)" if empty.

6. Display Change Log (execute.change_log[]). For each entry:

   | Entry ID | Plan Section | Files Touched | PA | EG | RA | Critic Verdict | |---|---|---|---|---|---|---|

   Where PA/EG/RA come from critic_dimensions.PA, critic_dimensions.EG, critic_dimensions.RA (show "pending" if absent). Display "(no changes yet)" if empty.

7. Display Scope Amendments (execute.scope_amendments[]):

   | SA ID | Gate ID | Amendment File | Reason | Approved At | |---|---|---|---|---|

   Or "(none)" if empty.

8. Display Rollback Log (execute.rollback_log[]):

   | Entry ID | Gate ID | Reason | Rolled Back At | |---|---|---|---|

   Or "(none)" if empty.

9. Display Test Manifest (execute.test_manifest[]). For each entry:

   | Source File | Command | Last Result | Last Run At | |---|---|---|---|

   Read test-results.jsonl from the instance directory if present. For each manifest entry, match the most recent test-results.jsonl record whose covering_files[0] equals the entry's source_file. Show the matched record's exit_code (pass/fail) and timestamp in Last Result and Last Run At. Truncate test_command to ~40 chars for display. Display "(not yet populated)" if execute.test_manifest is empty.

10. Display Environment Attestations (execute.env_attestations[]):

    | Env | Status | Attested At | Attested By | |---|---|---|---|

    Or "(not yet attested)" if empty.

11. Call TaskList to get live task status. Display tasks grouped by status (in_progress, pending, completed) with owner and metadata.bar_id where available.

12. Display Recent Discoveries (last 5 entries from discoveries.jsonl):

    | Discovery ID | Type | Detected By | Proposed Outcome | Resolution | |---|---|---|---|---|

    Read discoveries.jsonl as newline-delimited JSON. Show "(none)" if file absent or empty. Show resolution as "(open)" if null or absent.

13. Read the last 50 lines of ${CLAUDE_PROJECT_DIR:-$(pwd -P)}/.claude/deepwork/<id>/log.md and display as Recent Activity.

14. If execute.plan_drift_detected is true, close with a highlighted notice:

    PLAN DRIFT DETECTED
    Plan hash stored in state does not match plan_ref on disk.
    Run /deepwork-execute-amend <gate-id> to re-verdict the affected gate before resuming.
    

Format everything as a clean, readable dashboard. Use markdown tables; keep output scannable.