ai-sdlc-workflow: Command for Claude Code

Module Development Loop

Act as Orchestrator for module: {feature_name}

If PLAN_APPROVED exists for this module → skip Phase 1–2.

Context hygiene

Do not open into context:

• .cursor/state/workflow-state.json (use CLI below for learning counter only)
• .cursor/state/module-*-loop.json
• .aisdlc/*

Resume from docs/plans/, docs/module-deps.md, docs/memory/*.

Learning counter (write-only side effect, stdout is one compact JSON line):

python3 .cursor/scripts/learning-counter.py get
python3 .cursor/scripts/learning-counter.py inc --module {feature_name}

Office UI (when .aisdlc/ exists):

python3 .cursor/scripts/office-event.py --agent <id> --status working|done|idle|waiting|error --task "…" --phase <phase> --workflow {feature_name}

---

Phase 0 — Orient

cat docs/memory/decisions.md docs/memory/gotchas.md docs/memory/shortcuts.md 2>/dev/null || true
cat docs/module-deps.md 2>/dev/null || true
python3 .cursor/scripts/office-event.py --agent architect-planner --status working --task "Orient {feature_name}" --phase restore --workflow {feature_name}

---

Phases 1–5

Same as before: brainstorm → plan → scaffold → execute → test → judge/fix.
Emit office-event.py on each phase enter/exit. Use context-builder.py per agent.
Hard-gate execute on docs/module-deps.md.

---

Phase 6 — Done + Memory + Learning counter

1. Append docs/retrospective.md
2. Facts → docs/memory/* (1–5)
3. docs/module-deps.md → module done
4. Increment counter (do not edit state JSON by hand):

python3 .cursor/scripts/learning-counter.py inc --module {feature_name}
# stdout e.g. {"modulesSinceLastProposal": 3, "fullPassRecommended": false}

1. Dispatch @learning-agent:
   • If stdout fullPassRecommended: true → task text includes full pass
   • Else → lightweight scan

python3 .cursor/scripts/office-event.py --agent learning-agent --status working --task "Skill scan after {feature_name}" --phase review --workflow {feature_name}
python3 .cursor/context/context-builder.py \
  --phase review \
  --task "post-module skill scan for {feature_name}" \
  --agent learning-agent \
  --keywords "retrospective,pattern,skill,learning,gotcha,shortcut" \
  --budget 5000
python3 .cursor/scripts/office-event.py --agent learning-agent --status done --task "Learning finished" --phase review --workflow {feature_name}