superpowers-learning-workflow

Superpowers Learning Workflow

Overview

Use this workflow after meaningful work to capture what should survive the current session. It is a lightweight, repo-owned learning loop inspired by reflective agent systems, but scoped for safe use inside normal project workflows.

This is an explicit opt-in workflow. Do not use it by default. Only use it when the user explicitly asks for this workflow, names this skill, or a repository policy explicitly requires it.

Workflow

1. Review the recent work, decisions, and verification evidence.
2. Classify what was learned into four buckets:
   • durable project facts
   • current working state
   • session outcome
   • reusable method or repeated pitfall
3. Add required metadata for durable entries:
   • id
   • status
   • confidence
   • source
   • last_updated
   • review_after Do not mark an entry as verified if source is empty.
4. If .superpowers-memory/ exists, update:
   • PROJECT_CONTEXT.md for durable facts
   • CURRENT_STATE.md for active state
   • DECISIONS.md for lasting decisions
   • KNOWN_FAILURES.md for repeated failure patterns
   • VERIFICATION_BASELINE.md for trusted verification rules
   • TEAM_PREFERENCES.md for durable team agreements
   • USER_PROFILE.md for durable user preferences that are not project facts
   • AGENT_NOTES.md for durable execution reminders that are not project facts
   • session-journal/ for the session summary
   • LEARNING_BACKLOG.md for reusable patterns that may deserve future workflows or skills
5. If .superpowers-memory/ does not exist, tell the user to install the memory scaffold or keep the learning summary in a normal project doc.
6. Check whether any backlog item is strong enough to recommend promotion into a checklist, project rule, workflow step, script, or skill draft.
7. Review .superpowers-memory/SESSION_CLOSE_CHECKLIST.md before finishing the learning capture.
8. Use scripts/suggest-superpowers-memory-updates.ps1 if it is unclear which memory surfaces should be updated from the current session signals.
9. Prefer scripts/run-superpowers-memory-closeout.ps1 as the standard closeout helper when you want one command to review the checklist, get update suggestions, and optionally run validation.
10. When memory files were updated, run scripts/validate-superpowers-memory.ps1 and include the result in the summary.
11. Use scripts/search-superpowers-memory.ps1 when you need to confirm whether a pattern already exists in durable memory or recent journals.
12. Summarize what was learned and what, if anything, should become a future rule, checklist, script, or skill.