session-wrap

Session Wrap

This skill wraps up the current coding session, ensuring no context is lost and the handoff to the next session is seamless.

Workflow

1. Commit and PR: Create a PR for any work not already committed to a PR. If you need to use git, ensure you confirm the repo and completion.
2. Log Session: Update the session log (docs/session-logs or similar). Ensure you have per-session documentation for all previous sessions for this repo. Use agentic orchestration to explore and update as needed.
3. Prepare Next Session: Update next-session.md and prepare for the next session. Keep in mind there may have been another session running that just completed; incorporate its handoff points if applicable.
4. Update Claude Learnings: Use agentic orchestration to explore the session logs. Update CLAUDE.md with any information from the agents that might be pertinent to future sessions, especially repetitive errors and noticed problems that affect progress. We want to make sure future sessions improve upon previous ones. Update CLAUDE.md once findings are prepared.
5. Finalize: Confirm completion once all the logs and PRs are updated.

Evidence-First Handoff Protocol

Adapted from Agent33's evidence capture framework. Every session wrap must include verifiable evidence so the next session can trust previous work without re-running it.

Minimum Evidence Requirements Per Session

Every session wrap must capture at minimum:

1. Commands run: Exact CLI commands with their output summaries and exit codes
2. Test results: Test suite name, pass/fail counts, and coverage (if available)
3. Diff summary: Files changed with lines added/removed and rationale for each
4. Review outcomes: Reviewer feedback, required issues, and approval status (if applicable)

Evidence Capture Format

Use this template in the session log to record evidence:

## Session Evidence

### Commands Run
| Command | Output Summary | Exit Code |
|---------|---------------|-----------|
| `npm test` | 24 passed, 0 failed | 0 |
| `npm run lint` | No errors | 0 |

### Full Output (if non-trivial)
<paste full output for audit trail>

### Diff Summary
| Files Changed | Lines Added | Lines Removed | Rationale |
|---------------|-------------|---------------|-----------|
| `src/parser.ts` | +15 | -3 | Added error handling for malformed input |
| `tests/parser.test.ts` | +20 | -0 | Added test for edge case |

### Test Results
- Test suite: unit tests
- Outcome: 24 passed, 0 failed
- Coverage: 87% lines
- Notes: All acceptance criteria covered

### Evidence Checklist
- [ ] Commands recorded with exact CLI
- [ ] Outputs captured (summary + full if non-trivial)
- [ ] Diff summary documented
- [ ] Tests recorded with pass/fail count
- [ ] Review outcomes captured (if applicable)
- [ ] Artifacts linked (logs, reports, screenshots)