| name | retro |
| tier | capability |
| runtime | manual |
| trigger | mention |
| description | Engineering retrospective from git history: analyzes commit patterns, type mix (feat/fix/refactor), focus score, test discipline, and productivity trends. Use for Sprint/weekly reviews. Users may say: "retro", "retrospective", "sprint summary", "what did I do this week", "work analysis", "复盘", "回顾", "Sprint总结", "这周做了什么", "工作分析". Do NOT use for session wrap-up (use /handoff) or code review (use /review). |
| metadata | {"source":"oh-my-dag","version":"1.0.0","methodology":"git analytics retrospective"} |
| disable-model-invocation | true |
/retro — Engineering Retrospective
Auto-generate an engineering retrospective from git history. Not from memory — from data.
Input
/retro → default: engineering retro over the last 7 days of git
/retro 24h → last 24 hours
/retro 14d → last 14 days
/retro 30d → last 30 days
/retro compare → this week vs last week comparison
Workflow
Step 1: Data collection (1 message, 3 parallel Bash)
git log --since="{window_start}" --format="%H|%aN|%ai|%s" --shortstat
git log --since="{window_start}" --name-only --format="" | sort | uniq -c | sort -rn | head -20
git log --since="{window_start}" --format="%at|%ai|%s" | sort -n
Step 2: Analysis (pure reasoning)
2a. Commit Type Mix
Classify by the type prefix of each commit message:
| Type | Meaning | Health indicator |
|---|
feat | new feature | primary output |
fix | fix | < 30% is healthy |
refactor | refactor | deliberate improvement |
chore | chore | necessary but non-core |
docs | docs | knowledge capture |
perf | performance | optimization investment |
test | tests | quality investment |
Health signal: feat > 40% + fix < 30% = building mode. fix > 50% = firefighting mode.
2b. Focus Score
focus = 1 - (unique_directories_touched / total_commits)
| Score | Meaning |
|---|
| > 0.7 | highly concentrated (good: deep work) |
| 0.4-0.7 | moderately spread (normal: multitasking) |
| < 0.4 | over-spread (warning: too much context switching) |
2c. Session Detection
Commit gap > 2 hours → new session. Stats:
- Session count / average duration
- Peak hours (the densest commit window)
- Longest continuous work stretch
2d. Hotspots
File change frequency top 10. High-frequency files = potential instability zones or active development zones.
2e. Sprint Progress Check (if a progress map exists)
If your project keeps a sprint/progress map, read it → compare this cycle's plan vs actual completion.
Step 3: Output report
## Engineering Retro — {window} ({start} ~ {end})
### Summary
{1-2 sentences: the biggest achievement of this cycle}
Commits: {N} | Sessions: {M} | Focus: {score}
### Commit Mix
| Type | Count | % | Trend |
|------|-------|---|-------|
feat/fix/refactor/...
### Health Signals
- {building mode | firefighting mode | balanced}
- Fix share: {N}% {healthy | watch | warning}
- Test share: {N}% {sufficient | insufficient}
### Hotspots (Top 5 Most Changed Files)
| File | Changes | Type |
|------|---------|------|
### Session Patterns
- Peak hours: {HH:MM - HH:MM}
- Avg session: {N}h {M}min
- Longest session: {X}h
### Sprint Progress
- Planned: {items}
- Completed: {items} ({N}%)
- Blocked: {items}
### Retrospective
- **Keep**: {1-2 things that went well, anchored in data}
- **Improve**: {1-2 things to do differently, with specific suggestion}
- **Try**: {1 experiment for next sprint}
Step 4: Compare mode (/retro compare)
Compare two windows side by side:
| Metric | This Week | Last Week | Delta |
|--------|-----------|-----------|-------|
| Commits | ... | ... | +/- |
| feat% | ... | ... | +/- |
| fix% | ... | ... | +/- |
| Focus | ... | ... | +/- |
| Sessions | ... | ... | +/- |
Constraints
- Read-only — does not modify any file
- Output goes directly to the user, no file written
- Use
origin/ branches for git queries (local main may be stale)
- Use the user's local timezone for timestamps
- 0 commits in the window → report and suggest a different window
- Self-contained skill — does not read other docs
