What does this skill do, and when should you use it?
Git Workflow and Versioning is one skill in the addyosmani/agent-skills collection, focused on code changes, branches, commits, conflict handling, releases, and version management. It recommends short-lived feature branches integrated into a deployable main branch, with each verified increment committed as an atomic save point. It also covers worktrees, pre-commit hygiene, generated files, Git-based debugging, semantic versioning, release tags, and human-oriented changelogs. It fits teams that want AI coding agents to produce a clean, inspectable, and recoverable history.
Instructs the agent to use Git for every code change; recommends short-lived feature, fix, chore, or refactor branches from main; structures work as implement, test, verify, and commit increments; formats commit messages with a type and short description while explaining intent; uses git worktree for parallel branch work; checks staged diffs, secrets, tests, linting, and type checking before commits; distinguishes project-expected generated files from build output, environment files, and IDE settings; uses git bisect, log, diff, blame, and commit-message search for debugging; selects MAJOR, MINOR, or PATCH according to semantic versioning; creates release tags; maintains changelogs grouped by user impact; and produces structured post-change summaries covering changes made, untouched scope, and potential concerns.
- An AI agent is implementing a multi-file feature and needs tested slices with recoverable save points.
- Several agents are developing separate features in parallel and need isolated branches and directories through Git worktrees.
- A team is preparing a merge and needs checks for commit scope, message quality, secrets, tests, linting, and type safety.
- A maintainer is releasing software with consumers and needs a semantic version decision, an immutable tag, and a curated changelog.
- A developer is investigating a regression and needs Git commands to identify the introducing commit or inspect recent changes.
How do you install this skill?
- The unconditional “Always” trigger may over-apply to read-only analysis, documentation edits, or tasks without Git.
- Add explicit user confirmation, target validation, and recovery guidance before git reset --hard, worktree removal, or tag pushing.
- npm test, npm run lint, and npx tsc --noEmit are not present in every repository; detect the project toolchain first and explain failures.
- Chinese-language guidance and stable installation or reachability from mainland-China networks are not evidenced.
- Shell / CLI
- Network access
- Local filesystem
GitNode.js/npm
Install it from the repository with: npx skills add addyosmani/agent-skills --skill git-workflow-and-versioning. To browse the available skills first, run: npx skills add addyosmani/agent-skills --list. The README also documents installing the full collection of 24 skills with: npx skills add addyosmani/agent-skills.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/addyosmani/agent-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/git-workflow-and-versioning" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Apply git-workflow-and-versioning to this code change: split the work into tested atomic commits, check the staged diff for secrets, and recommend the version bump.
Activate it whenever you make code changes, commit, branch, resolve conflicts, coordinate parallel work, cut a release, choose a version bump, or write a changelog. Example prompt: "Apply git-workflow-and-versioning to this code change: split the work into tested atomic commits, check the staged diff for secrets, and recommend the version bump." The SKILL.md does not define a separate client-specific invocation syntax.
What are this skill's strengths and limitations?
- Covers the workflow from branches and commits through release tags, semantic versioning, and changelogs.
- Provides concrete branch names, commit formats, worktree commands, and pre-commit checks.
- Emphasizes atomic commits, short-lived branches, verification gates, and scope discipline for easier review, debugging, and rollback.
- Includes rationalization rebuttals and a practical list of red flags.
- It provides workflow guidance; it does not execute Git commands or inspect a project's actual state by itself.
- The npm test, npm run lint, and npx tsc --noEmit examples are not tailored to a particular project configuration.
- It does not document integration with a specific Git host, CI system, or release tool.
- The source provides no independent test suite or platform-compatibility test evidence for this individual skill.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Git Workflow & Release Discipline this page | 54 · Use with care | ★ 107k | 14d ago | MIT |
| Universal Release Flow | 51 · Use with care | ★ 4.4k | 1mo ago | MIT |
| Release Notes Drafting Skill | 32 · Not recommended | ★ 12k | 13d ago | MIT |
| DeepChat Release Assistant | 46 · Use with care | ★ 6.4k | 15d ago | Apache-2.0 |
| Bump Version — Automated Release Workflow for AionUi | 38 · Not recommended | ★ 34k | 14d ago | Apache-2.0 |
How did FollowSkills review this skill?
The skill promotes atomic commits, staged-diff inspection, secret checks, commit-based rollback, and worktree isolation, supporting a partial trust score. However, it applies to every code change without requiring user confirmation, and includes git reset --hard, worktree removal, and tag pushing without explicit consent, permission boundaries, sensitive-data guidance, or recovery safeguards; 9 points are deducted.
The instructions are internally organized across commits, branches, worktrees, releases, and verification, and repository CI validates skill structure and evaluation scripts. Reliability is limited because commands assume Node/npm/tsc and a particular project setup, offer little environment detection or abnormal-input handling, and provide thin failure diagnostics; static review cannot establish successful execution, so 11 points are deducted.
The audience and common scenarios are reasonably clear, covering code changes, parallel agents, releases, and versioning. The unconditional “Always” trigger risks false positives, while non-fit boundaries, input/output expectations, trigger exclusions, Chinese-language support, and mainland-China reachability are not specified; 7 points are deducted.
The skill has readable information architecture, examples, anti-rationalization guidance, red flags, and verification checklists. The README supplies installation guidance, and repository material states MIT licensing, a maintainer team, and CI. The skill lacks its own version, changelog, explicit maintenance owner, and update path, and assumes cross-skill references without documenting their availability; 5 points are deducted.
Atomic commits, short-lived branches, staged-diff checks, semantic versioning, and release checklists are directly usable for producing reviewable and reversible Git work. The guidance is prescriptive, offers limited adaptation for different repositories or non-Node projects, and includes heuristics such as the approximately 100-line target without task-specific justification; the static ceiling permits 7, with 8 points deducted for limited completeness and comparative evidence.
The skill includes concrete commands, checklists, and observable Git-state checks. Repository CI provides real workflow evidence for structural validation and evaluation scripts, supporting a limited score. The key Git behaviors are not covered by a committed skill-specific test suite, external claims such as DORA research are uncited, and no skill-specific evaluation results are supplied; 6 points are deducted.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →