| name | repo-hygiene |
| description | Periodic repository health check — dependencies, git, CI/CD, code quality, docs, security. Use when: onboarding to a repo, weekly maintenance, after big refactors, before audits, "is this repo in good shape?". Triggers: "repo hygiene", "health check", "repo health", "clean up repo", "maintenance check", "audit repo", "repo audit". |
repo-hygiene — Repository Health Check
A structured, periodic health check for repositories. Run anytime to detect drift,
accumulation of tech debt, and configuration issues before they become problems.
Not a release gate. For pre-release checks, use the pre-release skill instead.
This skill is for ongoing maintenance — run it weekly, after major refactors, when
onboarding to a repo, or whenever things feel "off".
When to Use
- Onboarding to a new (or forgotten) repo — "what shape is this in?"
- Weekly/monthly maintenance sweep
- After a large refactor or dependency upgrade
- Before starting a new feature sprint
- When CI starts failing mysteriously
- After a team member leaves and you inherit their repo
Supported Stacks
| Stack | Package manager | Detected by |
|---|
| Node.js / TypeScript | npm | package.json + package-lock.json |
| Python | uv / pip | pyproject.toml or requirements.txt |
| Go | go modules | go.mod |
Detect the stack from the project root. Multiple stacks in one repo is fine — run
applicable checks for each. If the stack isn't listed, skip stack-specific checks
and run the universal ones (git, CI, docs, security).
The Workflow
Step 0: Detect Project
ls package.json pyproject.toml go.mod 2>/dev/null
git rev-parse --show-toplevel
Determine: stack(s), git remote, default branch, CI system (GitHub Actions, GitLab CI, etc.).
Step 1: Dependency Health
Node.js / npm
| # | Check | Command | Severity |
|---|
| D1 | Known vulnerabilities | npm audit --json | 🔴 critical/high = Fix Now, moderate = Fix Soon |
| D2 | Outdated dependencies | npm outdated --json | 🟡 major bumps = Fix Soon, minor/patch = Info |
| D3 | Unused dependencies | npx depcheck --json | 🟡 Fix Soon |
| D4 | Phantom dependencies (used but undeclared) | npx depcheck --json → missing | 🔴 Fix Now |
| D5 | Lockfile freshness | See below | 🟡 Fix Soon |
| D6 | Duplicate dependencies | npm ls --all --json 2>/dev/null | grep -c '"deduped"' | ℹ️ Info |
D5 — Lockfile freshness check:
LOCK_DATE=$(git log -1 --format=%ct -- package-lock.json 2>/dev/null || echo 0)
PKG_DATE=$(git log -1 --format=%ct -- package.json 2>/dev/null || echo 0)
if [ "$PKG_DATE" -gt "$LOCK_DATE" ]; then
echo "⚠️ package.json modified after lockfile — run npm install"
fi
How to fix:
- D1:
npm audit fix for compatible fixes; npm audit fix --force for breaking (review changes). For stubborn advisories: check if the vuln is reachable, or override in package.json overrides.
- D2:
npm update for minor/patch; npm install <pkg>@latest for major (check changelogs).
- D3:
npm uninstall <pkg> for each unused dep.
- D4:
npm install <pkg> for each missing dep.
- D5:
npm install to regenerate lockfile, commit it.
- D6:
npm dedupe then verify tests pass.
Python (uv / pip)
| # | Check | Command | Severity |
|---|
| D1 | Known vulnerabilities | pip-audit --format=json | 🔴 Fix Now |
| D2 | Outdated dependencies | uv pip list --outdated or pip list --outdated --format=json | 🟡 Fix Soon |
| D3 | Unused dependencies | deptry . --json (if available) | 🟡 Fix Soon |
| D5 | Lockfile freshness | Compare uv.lock vs pyproject.toml timestamps | 🟡 Fix Soon |
How to fix:
- D1:
uv pip install --upgrade <pkg> for each vulnerable package. Check advisories for minimum safe version.
- D2:
uv pip install --upgrade <pkg> per package, or uv lock --upgrade for all.
- D3: Remove from
[project.dependencies] in pyproject.toml, then uv sync.
- D5:
uv lock && uv sync.
Go
| # | Check | Command | Severity |
|---|
| D1 | Known vulnerabilities | govulncheck ./... | 🔴 Fix Now |
| D2 | Outdated dependencies | go list -m -u all | 🟡 Fix Soon |
| D3 | Unused dependencies | go mod tidy -v (reports removed) | 🟡 Fix Soon |
How to fix:
- D1:
go get <module>@latest for vulnerable deps, then go mod tidy.
- D2:
go get -u ./... for all, or go get <module>@latest selectively.
- D3:
go mod tidy removes unused; commit go.mod and go.sum.
Step 2: Git Hygiene
| # | Check | Command | Severity |
|---|
| G1 | Stale local branches (merged) | git branch --merged main | grep -v '^\*|main|develop' | 🟡 Fix Soon |
| G2 | Stale remote branches (merged) | git branch -r --merged origin/main | grep -v 'HEAD|main|develop' | 🟡 Fix Soon |
| G3 | Large files in repo | See below | 🟡 Fix Soon (🔴 if >10MB) |
| G4 | .gitignore completeness | See below | 🟡 Fix Soon |
| G5 | Untracked files that should be ignored | git status --porcelain | grep '^??' — look for build artifacts, IDE files, env files | ℹ️ Info |
| G6 | Uncommitted changes | git status --porcelain | ℹ️ Info |
G3 — Large files check:
git ls-files -z | xargs -0 -I{} git log --diff-filter=A --format='%H' -1 -- '{}' | head -20
git ls-files -z | xargs -0 du -sh 2>/dev/null | sort -rh | head -10
G4 — .gitignore completeness:
Must include (per stack):
- Universal:
.env, .env.*, *.local, .DS_Store, Thumbs.db, *.swp, .idea/, .vscode/ (or be deliberate about tracking it)
- Node:
node_modules/, dist/, build/, coverage/, .turbo/, .next/
- Python:
__pycache__/, *.pyc, .venv/, venv/, .mypy_cache/, .pytest_cache/, *.egg-info/
- Go: binary name (check
go build -o), vendor/ (if not vendoring)
How to fix:
- G1:
git branch -d <branch> for each merged local branch.
- G2:
git push origin --delete <branch> for each merged remote branch. Be careful — confirm with team.
- G3: For files that shouldn't be tracked: add to
.gitignore, git rm --cached <file>. For files already in history: git filter-repo or BFG Repo-Cleaner (destructive — confirm first).
- G4: Add missing patterns to
.gitignore. Use a generator like gitignore.io as a starting point.
- G5: Either add to
.gitignore or git add if they should be tracked.
Step 3: CI/CD Health
Skip if no CI configuration found.
| # | Check | Command | Severity |
|---|
| C1 | Workflow files exist | ls .github/workflows/*.yml 2>/dev/null | ℹ️ Info |
| C2 | Actions pinned by SHA | grep -rE 'uses: [^@]+@v[0-9]' .github/workflows/ — should return nothing | 🟡 Fix Soon |
| C3 | Least-privilege permissions | Scan for permissions: blocks; flag write-all or missing job-level perms | 🟡 Fix Soon |
| C4 | No secret leaks in workflows | Check for echo ${{ secrets.* }}, secret in $GITHUB_OUTPUT/$GITHUB_ENV | 🔴 Fix Now |
| C5 | Deprecated actions | Check for known deprecated: actions/create-release@v1, set-output commands, ::set-env | 🟡 Fix Soon |
| C6 | Node/Python version matches project | Compare workflow matrix with engines, .nvmrc, pyproject.toml [requires-python] | 🟡 Fix Soon |
How to fix:
- C2: Replace
uses: actions/checkout@v4 with uses: actions/checkout@<full-sha>. Find SHA: gh api repos/actions/checkout/git/ref/tags/v4 --jq .object.sha or check the releases page.
- C3: Add explicit
permissions: at job level. Start with contents: read and add only what's needed.
- C4: Remove secret interpolation. Use
environment: blocks or write to files with masking.
- C5: Replace deprecated actions with current equivalents.
set-output → $GITHUB_OUTPUT file.
- C6: Align versions. Use
.nvmrc or engines as the source of truth.
Step 4: Code Quality Drift
| # | Check | Command | Severity |
|---|
| Q1 | TODO/FIXME/HACK count | git grep -ciE '(TODO|FIXME|HACK)' -- '*.ts' '*.js' '*.py' '*.go' ':!node_modules' ':!vendor' ':!.venv' | ℹ️ Info (🟡 if >20) |
| Q2 | console.log in src (JS/TS) | git grep -c 'console\.log' -- 'src/**/*.ts' 'src/**/*.js' ':!*.test.*' ':!*.spec.*' | 🟡 Fix Soon |
| Q3 | Disabled/skipped tests | git grep -cE '(it\.skip|test\.skip|describe\.skip|xit|xdescribe|@pytest\.mark\.skip|t\.Skip)' -- '*.test.*' '*.spec.*' '*_test.*' '*_test.go' | 🟡 Fix Soon |
| Q4 | Lint passes | npm run lint / ruff check . / golangci-lint run | 🟡 Fix Soon |
| Q5 | Tests pass | npm test / pytest / go test ./... | 🔴 Fix Now |
| Q6 | Build succeeds | npm run build / uv build / go build ./... | 🔴 Fix Now |
| Q7 | Type errors (TS) | npx tsc --noEmit | 🟡 Fix Soon |
| Q8 | Dead exports (TS) | npx ts-prune 2>/dev/null | grep -v '(used in module)' | ℹ️ Info |
How to fix:
- Q1: Triage each TODO — either do it, create an issue/task for it, or remove it if obsolete.
- Q2: Replace with a proper logger, or remove debug logging.
grep -rn 'console.log' src/ to find them.
- Q3: Either fix the underlying issue and un-skip, or delete the test if the feature was removed.
- Q4–Q7: Fix the errors. Run the tool, address each issue.
- Q8: Remove unused exports, or add
// ts-prune-ignore-next if they're part of the public API.
Step 5: Documentation Freshness
| # | Check | Command | Severity |
|---|
| F1 | README.md exists | File check | 🔴 Fix Now |
| F2 | README freshness vs. source | Compare git log -1 --format=%cr -- README.md vs git log -1 --format=%cr -- src/ | 🟡 if src is >30 days newer |
| F3 | CHANGELOG exists | File check | 🟡 Fix Soon (for published packages) |
| F4 | Broken internal links | See below | 🟡 Fix Soon |
| F5 | LICENSE file present | File check | 🔴 Fix Now |
| F6 | LICENSE matches package metadata | Compare LICENSE text with package.json license / pyproject.toml license | 🟡 Fix Soon |
| F7 | AGENTS.md references valid paths | If .pi/AGENTS.md exists, check that referenced files/dirs exist | 🟡 Fix Soon |
F4 — Broken link check:
grep -roE '\[([^]]+)\]\(([^)]+)\)' *.md docs/**/*.md 2>/dev/null | \
grep -v 'http' | \
while IFS= read -r line; do
path=$(echo "$line" | sed 's/.*](\([^)]*\)).*/\1/' | sed 's/#.*//')
if [ -n "$path" ] && [ ! -e "$path" ]; then
echo "BROKEN: $line"
fi
done
How to fix:
- F1: Write a README with: what it does, how to install, how to use, prerequisites, license.
- F2: Review README against current code — update examples, API docs, feature lists.
- F3: Add a CHANGELOG.md. Consider
@changesets/cli for automated generation (see pre-release skill).
- F4: Update or remove broken links.
- F5: Add a LICENSE file. Use choosealicense.com if unsure.
- F6: Make LICENSE file and metadata agree.
- F7: Update AGENTS.md to reflect current project structure.
Step 6: Configuration Consistency
| # | Check | How | Severity |
|---|
| X1 | EditorConfig present | .editorconfig exists | ℹ️ Info |
| X2 | Strict mode (TS) | tsconfig.json → "strict": true | 🟡 Fix Soon |
| X3 | Formatter configured | .prettierrc / ruff.toml / gofmt (built-in) | 🟡 Fix Soon |
| X4 | Linter configured | .eslintrc* or eslint.config.* / ruff.toml / golangci-lint config | 🟡 Fix Soon |
| X5 | Engine constraints match CI | package.json engines vs CI matrix; pyproject.toml requires-python vs CI | 🟡 Fix Soon |
| X6 | .nvmrc / .python-version matches | Compare with engines / requires-python / CI config | ℹ️ Info |
How to fix:
- X1: Add
.editorconfig. Minimal: root = true, [*] block with indent_style, indent_size, end_of_line, insert_final_newline.
- X2: Set
"strict": true in tsconfig.json. Fix resulting type errors (usually worth it).
- X3–X4: Add config files. Use the project's existing style as a baseline.
- X5–X6: Pick one source of truth (recommend
engines / requires-python) and align everything else.
Step 7: Security Posture
Lightweight security checks for ongoing hygiene. For the full pre-release security audit
(gitleaks, trufflehog, workflow audit), use the pre-release skill.
| # | Check | Command | Severity |
|---|
| S1 | No tracked .env or .local files | git ls-files '*.env' '*.env.*' '*.local' '*.local.*' '.env' '.env.local' | 🔴 Fix Now |
| S2 | .env.example exists (if .env in .gitignore) | File check | 🟡 Fix Soon |
| S3 | No hardcoded secrets in source | git grep -iE '(api[_-]?key|secret|password|token)\s*[:=]\s*["\x27][^"\x27]{8,}' -- ':!*.lock' ':!node_modules' ':!*.example' ':!*.sample' | 🔴 Fix Now |
| S4 | Secrets scanning config present | .gitleaks.toml or pre-commit hooks | ℹ️ Info |
| S5 | No broad file permissions | Check for chmod 777 or 0777 in scripts | 🔴 Fix Now |
How to fix:
- S1:
git rm --cached <file>, add to .gitignore, commit. If the file contained real secrets, rotate them immediately — they're in git history.
- S2: Create
.env.example with placeholder values (<REPLACE_ME>) for every var in .env.
- S3: Move secrets to env vars or a secrets manager. Replace in code with
process.env.VAR / os.environ["VAR"].
- S4: Add
.gitleaks.toml (even a minimal one enables CI scanning). Or add gitleaks to pre-commit hooks.
- S5: Use least-privilege permissions (
644 for files, 755 for executables).
Step 8: Project Metadata
| # | Check | How | Severity |
|---|
| M1 | Required package fields | name, version, description, license in package.json / pyproject.toml | 🟡 Fix Soon |
| M2 | Repository URL set | repository field in package metadata | 🟡 Fix Soon |
| M3 | Keywords present | keywords array | ℹ️ Info |
| M4 | FUNDING.yml (public repos) | .github/FUNDING.yml exists | ℹ️ Info |
| M5 | Pi package compliance | If ships skills/extensions: pi-package keyword, pi manifest, files includes skill dirs | 🟡 Fix Soon (if applicable) |
How to fix:
- M1–M3: Add the missing fields to
package.json or pyproject.toml.
- M4: Create
.github/FUNDING.yml with github: <username>.
- M5: See pi package docs for required fields.
Baseline Tracking
Save a baseline after each run to detect drift over time. Store at .pi/hygiene-baseline.json:
{
"timestamp": "2026-02-14T23:00:00Z",
"stack": ["node"],
"scores": {
"dependencies": { "status": "healthy", "vulns": 0, "outdated": 3, "unused": 0 },
"git": { "status": "healthy", "stale_branches": 0, "large_files": 0 },
"ci": { "status": "warning", "unpinned_actions": 2, "permission_issues": 0 },
"quality": { "status": "healthy", "todos": 5, "skipped_tests": 0, "lint_clean": true },
"docs": { "status": "warning", "readme_stale_days": 45, "broken_links": 1 },
"config": { "status": "healthy", "strict_ts": true, "formatter": true, "linter": true },
"security": { "status": "healthy", "tracked_env": 0, "hardcoded_secrets": 0 },
"metadata": { "status": "healthy", "complete": true }
},
"overall": "7/10"
}
On subsequent runs, compare with baseline and flag regressions:
📉 Dependencies: 0 → 3 vulnerabilities (regression since last check)
📈 Quality: 15 → 5 TODOs (improvement!)
→ CI: unchanged — 2 unpinned actions remain
When the user approves the report, offer to update the baseline.
Report Format
Present the final report as a health scorecard:
# Repo Health: <project-name>
## Score: 7/10 — GOOD
## Stack: Node.js + TypeScript
## Last check: 2026-01-15 (30 days ago) | Baseline: 6/10 📈
### 🔴 Fix Now (2)
| # | Category | Issue | Fix |
|---|----------|-------|-----|
| S1 | Security | `.env.local` tracked in git | `git rm --cached .env.local` |
| D1 | Deps | 2 high-severity npm audit findings | `npm audit fix` |
### 🟡 Fix Soon (4)
| # | Category | Issue | Fix |
|---|----------|-------|-----|
| D2 | Deps | 8 outdated packages (2 major) | `npm outdated` → upgrade |
| C2 | CI | 3 actions not pinned by SHA | Pin to commit SHA |
| F2 | Docs | README 45 days behind source | Review and update |
| Q3 | Quality | 2 skipped tests | Fix or remove |
### 🟢 Healthy (12)
- ✅ Dependencies: no unused, no phantom, lockfile fresh
- ✅ Git: clean tree, no stale branches, no large files
- ✅ Code: lint clean, build passes, tests pass, strict TS
- ✅ Config: EditorConfig, Prettier, ESLint all configured
- ✅ Security: no tracked secrets, .env.example present
- ✅ Metadata: all fields present, license matches
### 📊 Trends (vs. baseline 2026-01-15)
| Category | Then | Now | Trend |
|----------|------|-----|-------|
| Vulnerabilities | 0 | 2 | 📉 |
| Outdated deps | 5 | 8 | 📉 |
| TODOs | 15 | 8 | 📈 |
| Skipped tests | 0 | 2 | 📉 |
### Recommendations
1. **Immediate**: Fix the 2 security/vulnerability items above
2. **This week**: Pin CI actions and update stale README
3. **Ongoing**: Address skipped tests and outdated deps in next sprint
Use your project's task tracking to schedule these items.
Scoring
Calculate the score from check results:
| Result | Points deducted |
|---|
| Each 🔴 Fix Now | −1.5 |
| Each 🟡 Fix Soon | −0.5 |
| ℹ️ Info | 0 |
Start at 10, apply deductions, floor at 0. Round to nearest integer.
| Score | Label |
|---|
| 9–10 | 🟢 EXCELLENT |
| 7–8 | 🟢 GOOD |
| 5–6 | 🟡 FAIR |
| 3–4 | 🟠 NEEDS WORK |
| 0–2 | 🔴 POOR |
Auto-Fix Offers
After presenting the report, offer to fix issues that are safe and mechanical.
Always present what will be done and get confirmation before executing.
Safe to offer (low risk, reversible)
- Delete merged local branches (
git branch -d)
- Run
npm audit fix (compatible fixes only, not --force)
- Run
npm dedupe / go mod tidy
- Add missing
.gitignore patterns
- Add
.editorconfig from template
- Remove
console.log from source files
- Create
.env.example from .env (with values replaced by <REPLACE_ME>)
- Add missing
package.json fields (description, repository, keywords)
- Create
.github/FUNDING.yml
Offer with warning (confirm carefully)
- Delete merged remote branches (
git push origin --delete)
- Run
npm audit fix --force (may have breaking changes)
- Major dependency upgrades
- Enable TypeScript strict mode (may produce many errors)
- Update CI action pinning (must verify correct SHAs)
Never auto-fix (explain, let user decide)
- Removing tracked
.env files (may need secret rotation)
- Rewriting git history (BFG / filter-repo)
- Changing license files
- Modifying CI permissions model
- Removing hardcoded secrets (need to determine replacement strategy)
Tips
- Run early, run often. A monthly cadence catches drift before it compounds.
- Don't try to fix everything at once. Focus on 🔴 items first, batch 🟡 items into a maintenance sprint.
- Baseline tracking is your friend. Even if the score isn't perfect, trending upward means you're winning.
- Pair with pre-release. Run
repo-hygiene for ongoing health, pre-release when you're ready to ship. They complement each other — hygiene keeps the baseline high so pre-release has fewer surprises.
- New repos start clean. Run this right after
git init to establish a perfect baseline. It's easier to maintain 10/10 than to recover from 4/10.