| name | resume-engagement |
| description | Resume from engagement_state.json. Use when continue client, pick up books, or where we left off.
|
/resume-engagement
Purpose
Make multi-session work safe. Disk is truth.
Workflow
- Find
engagement_state.json:
- cwd, or
- cwd/clients/*/engagement_state.json, or
- path user provides, or
- fixtures path if testing
- Load state +
shared/agent-runtime.md + shared/operator-lens.md + references/stage_artifacts.md
- If
operator missing → resolve once and write it (do not re-ask if inferable)
- Verify artifacts listed in
state.artifacts still exist
- Run
python3 scripts/validate_engagement_artifacts.py <client_dir> if workpapers present
- Print status board:
## Resume — [legal_name] · [operator] · [engagement_type] · period [fy / on-disk]
- Stage: [current_stage] ([status])
- Completed: [stages_completed]
- Blockers: [blockers or none]
- Open queries: [n]
- Next: [concrete next skill/action]
- If
status == blocked → work the blocker first
- If
status == waiting_on_user and open_queries non-empty →
re-issue unanswered asks via structured user-question tool
(shared/user-questions.md) — do not only re-print queries.md
- Else continue from
current_stage by loading that stage’s SKILL.md
- Do not restart the pipeline from setup unless user says so
If no state file
Say so clearly. Offer:
engagement-setup for a new client
full-engagement-pipeline if they dumped files
Completion
Done when: state loaded, one-line status shown, next stage skill identified (or blockers listed).