Agent skill
npm-update-report
Dependency update workflow with vulnerability assessment and verification. Use when asked to check outdated packages, bump dependencies, run audit, do periodic dependency maintenance, or when security vulnerabilities are reported.
Install this agent skill to your Project
npx add-skill https://github.com/hairihou/dotfiles/tree/main/src/.config/claude/skills/npm-update-report
SKILL.md
Package Update Report
Context
- Lock files: !
ls package-lock.json pnpm-lock.yaml yarn.lock 2>/dev/null - Available scripts: !
node -e "const p=require('./package.json');console.log(Object.keys(p.scripts||{}).join(', '))" 2>/dev/null
Workflow
| Step | Action | Details |
|---|---|---|
| 1 | Detect PM | Identify PM from lockfile (see command table below) |
| 2 | Check outdated | List packages with available updates |
| 3 | Update packages | Update according to strategy below |
| 4 | Classify changes | Extract diff from package.json, classify as major/minor/patch |
| 5 | Investigate | Web search changelogs for major/minor bumps and key packages |
| 6 | Assess impact | Grep for package usage, evaluate breaking changes |
| 7 | Audit | Run security audit, include advisory URLs for vulnerabilities |
| 8 | Verify | Run python scripts/run-verification.py <pm> — outputs PASS/FAIL/SKIP per script |
| 9 | Output | Follow report template in ${CLAUDE_SKILL_DIR}/references/report-template.md |
Package Manager Detection
Detect PM from lockfile per CLAUDE.md rules. If no lockfile found, stop and inform user.
Monorepo workspace commands: pnpm --filter {pkg}, npm -w {pkg}, yarn workspace {pkg}
Update Strategy
| Type | Action |
|---|---|
| Patch | Auto-update via {pm} update |
| Minor | Auto-update, investigate key packages |
| Major | Confirm with user before update |
Investigation Criteria
Sources: GitHub Releases, CHANGELOG.md, official blogs only
Always investigate:
- Major version bumps (breaking changes likely)
- Minor bumps of: frameworks (React, Vue, Next.js), build tools (Vite, esbuild), test tools (Vitest, Jest)
Investigate if verification fails:
- Any package that may be related to the failure
Skip: Patch-only updates with passing verification
Error Handling
- No lockfile found → stop, inform user to run package manager install first
outdatedcommand returns empty → report "all packages up to date", skip remaining steps- Verification script not found (e.g., no
testscript) → skip that check, note in report - Major update causes verification failure → revert that specific package, note in report as "requires manual migration"
Verification Failure Handling
If verification fails:
- Identify failing script and error message
- Search for related packages in the error
- Investigate those packages' changelogs for breaking changes
- Document findings in report under "Verification Results"
- Set conclusion to "Needs attention" with specific action items
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
retrospect
Guided personal reflection that separates fact, interpretation, and emotion through structured inquiry. Use when "something feels off", "need to process this", "had a conflict", or decisions feel unclear. Not for technical decision analysis — use devils-advocate for that.
1on1-prep
Prepare for 1on1 meetings using a Psychological Safety x Responsibility framework. Use when preparing for a 1on1, noticing behavioral changes, or planning a difficult conversation with a report.
bonsai
Maintain and groom config files in the current repository. Use when "bonsai", "tidy up", "prune configs", "clean up", "spring cleaning", or reviewing config health. Not for code quality review — use code-critic for that.
session-insights
Analyze the current conversation for inefficiencies and improvement opportunities. Use when "how did this session go", "what went wrong", "what could I have done better", or reviewing session quality. Not for personal reflection — use retrospect. Not for logging work — use diary.
pr
Create or update a GitHub pull request. Use when asked to create a PR, submit for review, or push changes as a PR. Not for code review comments — use conventional-comments for that.
conventional-comments
Format code review comments using Conventional Comments labels and decorations. Use when writing inline review comments on a PR diff. Not for creating PRs — use pr for that. Not for complexity analysis — use code-critic for that.
Didn't find tool you were looking for?