agentleFS
jars-demojars-demo/jars-skills/skills/dependency-update/SKILL.md

dependency-update

Skill0 starsChanged 32 days agoMITView on GitHub

Audit outdated dependencies, review changelogs for breaking changes, update packages safely, and write a clean commit. Use when the user asks to update dependencies, upgrade packages, check for outdated deps, or bump a library version. Never upgrades a major version blindly — always reads the changelog first and surfaces breaking changes before touching any file.

  • Installs packages

10 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    One package at a time for risky upgrades: Don't batch a major + two minors into one commit.

    line 18
  2. Distinctive:

    Never skip hooks or force-write lockfiles: Use the package manager's standard update command so integrity checks and resolution logic run …

    line 21
  3. Distinctive:

    Tests must pass after every update: If they don't, report the failure honestly — don't hide it, don't proceed with the next package.

    line 23
  4. Distinctive:

    After each install, verify the lockfile changed as expected:

    line 129
  5. Distinctive:

    Run the full check suite:

    line 137
  6. Distinctive:

    Read the error carefully — it is usually a removed or renamed API.

    line 155
  7. Distinctive:

    If the fix is non-trivial, stop and report to the user before continuing.

    line 157
  8. Distinctive:

    Use the repository's existing commit style if it differs from the examples above.

    line 184

“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Dependency Update

Work through a dependency upgrade systematically: find what's outdated, read what changed, update one layer at a time, verify the build still passes, and leave a clear commit. The goal is to land updates that don't break anything and that a reviewer can audit in 30 seconds.

Ground rules

  • Read the changelog before touching anything. For any major or minor bump, find and read the migration guide or release notes. Never upgrade based on the version number alone.
  • Major versions require explicit user confirmation. Flag the breaking changes you found, explain what migration work is likely needed, and wait for a go-ahead before updating.
  • One package at a time for risky upgrades. Don't batch a major + two minors into one commit. Group patch updates together; keep each significant upgrade isolated.
  • Never push. Stage, commit, and stop. The user decides when to push.
  • Never skip hooks or force-write lockfiles. Use the package manager's standard update command so integrity checks and resolution logic run normally.
  • Tests must pass after every update. If they don't, report the failure honestly — don't hide it, don't proceed with the next package.

Workflow

1. Identify the package manager and current state

# Node / npm
cat package.json
npm outdated

# Node / yarn
yarn outdated

# Node / pnpm
pnpm outdated

# Python / pip
pip list --outdated

# Python / uv
uv pip list --outdated

# Rust
cargo outdated          # requires cargo-outdated

# Ruby
bundle outdated

Also check:

  • .nvmrc, .node-version, .python-version — confirm you're on the right runtime
  • package.json "engines" field or pyproject.toml requires-python — version constraints
  • CI config (.github/workflows/, Jenkinsfile) — if it pins versions there too, note it

2. Triage the outdated list

Build a table before touching anything:

PackageCurrentLatestJumpCategory
express4.18.24.21.0patch/minorsafe to batch
eslint8.57.09.12.0majorneeds changelog review
axios1.6.81.7.4minorsafe to batch

Categories:

  • safe to batch — patch or minor bumps with no noted breaking changes
  • needs changelog review — any major bump, or a minor bump on a library known for sharp edges
  • skip — packages the user hasn't asked about, or locked by a peer-dependency

3. Read the changelog for each significant bump

For every major version bump (and minors on critical libraries):

# Find changelog on npm
npm info <package> homepage
npm info <package> repository.url

# Or fetch directly
curl -s https://raw.githubusercontent.com/<owner>/<repo>/main/CHANGELOG.md | head -200

Look for:

  • Removed APIs — will your codebase call something that no longer exists?
  • Renamed exports, config keys, or CLI flags
  • Peer dependency changes — does the new version require a different version of something else?
  • Node.js / Python minimum version bumps

Report what you found before proceeding with a major bump. Wait for the user to confirm.

4. Apply updates

Batch safe updates (patches and minors):

# npm
npm update                           # patch + minor updates within semver ranges
npx npm-check-updates -u            # also update the ranges in package.json if needed
npm install

# yarn
yarn upgrade --latest --pattern "*" # or specify packages

# pnpm
pnpm update

# Python / uv
uv pip install --upgrade <package1> <package2>

# Python / pip
pip install --upgrade <package1> <package2>
pip freeze > requirements.txt       # if using a flat requirements file

# Rust
cargo update

Individual major upgrade (confirmed by user):

npm install <package>@<new-major>
# then apply any migration steps found in step 3

After each install, verify the lockfile changed as expected:

git diff package-lock.json   # or yarn.lock / pnpm-lock.yaml / Cargo.lock

5. Verify the build

Run the full check suite:

# Node
npm run build   # or tsc --noEmit
npm test

# Python
python -m pytest
# or
uv run pytest

# Rust
cargo build
cargo test

If anything fails:

  1. Read the error carefully — it is usually a removed or renamed API.
  2. Fix the call site, not the version pin.
  3. If the fix is non-trivial, stop and report to the user before continuing.

6. Commit

Group related updates into logical commits:

chore(deps): update patch and minor dependencies

Bumped:
- axios 1.6.8 → 1.7.4
- eslint-plugin-react 7.34.3 → 7.35.1
- vitest 1.6.0 → 1.6.4

For a major upgrade, use a separate commit with more context:

chore(deps)!: upgrade eslint 8 → 9

Breaking changes from upstream:
- Flat config is now required; migrated eslint.config.js
- Removed --ext flag; now configured via ignores pattern

Ref: https://eslint.org/docs/latest/use/migrate-to-9.0.0

Use the repository's existing commit style if it differs from the examples above.

7. Report back

Tell the user:

  • What was updated and to which version
  • Any breaking changes found and how they were handled (or flagged for later)
  • Any packages you skipped and why
  • Any test failures still unresolved

Examples

WeakStrong
Run npm update and commit everythingTriage first, batch safe updates, isolate majors
Upgrade webpack 4 → 5 without reading the migration guideRead the webpack 5 migration guide first, note breaking changes, confirm with user
chore: bump depschore(deps): update axios 1.6.8 → 1.7.4, vitest 1.6.0 → 1.6.4
Hide a failing test after the upgradeReport the failure explicitly; fix or revert before committing

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.