Mode d’emploi
- Copiez la commande et examinez le paquet avant de l’installer dans votre environnement agent.
- Consultez le document original pour les outils, identifiants et environnements requis.
- Décrivez les entrées et le livrable souhaité ; comparez la sortie aux sources.
npx skills add https://github.com/basecamp/house-skills --skill git-activityAvant de commencer
- L’installation n’inclut ni abonnement au modèle, ni quota API, ni accès aux fichiers.
- Étoiles et installations reflètent l’activité, pas un test indépendant de qualité ou de sécurité.
Rendu depuis le dépôt source en conservant titres, exemples, code, tableaux, liens et images.
Git Activity Fetcher
Fetch git commit history from local repositories into per-day cached JSON atoms at ~/.cache/recap/git/{repo}/{YYYY-MM-DD}/log.json. These atoms feed the /recap orchestrator for progressive timescale synthesis.
Invocation
/recap:git-activity --repos coworker:~/Work/basecamp/coworker --since 2026-03-23
/recap:git-activity --repos coworker:~/Work/basecamp/coworker,house-skills:~/Work/basecamp/house-skills --since 2026-03-23 --until 2026-03-30
/recap:git-activity --repos coworker:~/Work/basecamp/coworker --since 2026-03-23 --reuse
/recap:git-activity --repos coworker:~/Work/basecamp/coworker --since 2026-03-23 --author "Jeremy Daer" --all-branchesContract
- Idempotent: same input produces same output, safe to re-run
- Day-cached: one JSON file per repo per day at
~/.cache/recap/git/{repo}/{YYYY-MM-DD}/log.json,
or under a scope directory when --author or --all-branches narrows or widens the log
- `--reuse`: skip fetch if cache exists and is marked complete
- No API calls: reads local git repos only
Quick Run
# 1. Determine date range
SINCE=$(date -d "7 days ago" +%Y-%m-%d)
UNTIL=$(date +%Y-%m-%d)
# 2. Run the fetcher
"$SKILL_DIR/scripts/git-activity.sh" \
--repos "coworker:$HOME/Work/basecamp/coworker" \
--since "$SINCE" --until "$UNTIL"
# 3. Verify cache
ls ~/.cache/recap/git/coworker/
cat ~/.cache/recap/git/coworker/$SINCE/log.json | jq '.metadata'Where $SKILL_DIR = directory containing this SKILL.md.
Output Format
Each log.json contains:
{
"repo": "coworker",
"date": "2026-03-24",
"commits": [
{
"hash": "abc123...",
"short_hash": "abc123",
"subject": "Add weekly digest skill",
"author": "Jeremy",
"date": "2026-03-24T10:30:00-05:00",
"body": "Extended commit message..."
}
],
"metadata": { "complete": true, "count": 3 }
}Arguments
| Argument | Required | Description |
|---|---|---|
--repos | Yes | Comma-separated name:path pairs |
--since | Yes | Start date (YYYY-MM-DD) |
--until | No | End date (default: today) |
--author | No | Keep only commits whose author matches this pattern (git log --author) |
--all-branches | No | Log every local and remote-tracking branch, not just the checked-out one |
--reuse | No | Skip fetch if cache exists and is complete |
By default the fetcher logs the checked-out branch and every author: the repo's activity. For one person's work, pass --author and --all-branches, so coworkers' commits drop out and commits on other branches, pushed or not, come in.
Cache Structure
~/.cache/recap/git/
coworker/
2026-03-24/log.json
2026-03-25/log.json
all-branches-author-jeremy-daer-189a21ec/
2026-03-24/log.json
house-skills/
2026-03-24/log.jsonA filtered fetch caches under its own scope directory, named from the filters, so --reuse never hands a filtered run the unfiltered day or the other way round. The name ends in a digest of the exact --author pattern: Jeff and ^Jeff$ match different commits, so they must not share a directory.
Day-sized atoms enable progressive aggregation: a weekly digest reads 7 daily atoms. A monthly digest reads ~30. Different timescales, same data.
Failure Modes
| Symptom | Cause | Fix |
|---|---|---|
| "not a git repo" | Path doesn't point to repo | Check --repos paths |
| Empty commits for a day | No commits on that date | Expected — empty days are cached as complete |
| awk/jq parse error | Unusual characters in commit messages | Falls back to simpler format automatically |


