weekly-reviewlisted
Install: claude install-skill tomcounsell/ai
# Weekly Review
Produces a structured engineering review of recent commits — N categories with 2-5 bullets each, plus contributor statistics — written in plain language a non-technical stakeholder can read while staying meaningful to engineers. Output is plain text with Unicode emojis, ready to paste into email, Slack, or a doc. Purely git-based, so it works in any codebase.
## Defaults
- `days`: 7 (use 14 for bi-weekly, 30 for monthly)
- `categories`: 5 (use 3 for shorter reviews, 7 for monthly)
If the user provides args, parse them as `<days> [categories]`. Otherwise use defaults.
## Phase 1: Gather data
Run these git commands in parallel from the repo root to collect commit history:
```bash
# Verify you're on the correct branch
git pull && git branch --show-current
# Get all commits
git log --since="<DAYS> days ago" --oneline --no-merges
# Count commits by author
git log --since="<DAYS> days ago" --format="%an" --no-merges | sort | uniq -c | sort -rn
# Get detailed stats (first 500 lines)
git log --since="<DAYS> days ago" --stat --no-merges | head -500
```
## Phase 2: Analyze internally (do not output this)
Think through the commits and organize them. Do NOT produce a long verbose breakdown — this is internal work.
1. **Review the commits** — read through and understand what changed
2. **Identify patterns** — group related commits together
3. **Choose N categories** — pick categories that naturally emerge from the actual work
4. **Note key stats** — total comm