agentleFS
mgiovanimgiovani/cc-arsenal/skills/gitflow/SKILL.md

gitflow

Skill8 starsChanged 48 days agoMITView on GitHub

Manage a gitflow branching workflow, starting and finishing feature,

  • Installs packages

6 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

6 rules. 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Read the matching reference in full and follow it step by step.

    line 32
  2. Distinctive:

    Conventional Commits: Follow the Conventional Commits spec.

    line 49
  3. Very specific:

    Never force-push: Never bypass git hooks (--no-verify or equivalent).

    line 50
  4. Very specific:

    Stage only the files you intend: Prefer git add <specific-files> over blanket staging commands.

    line 51
  5. Very specific:

    Verify before every push: Run git log --oneline <base>..HEAD and git diff --stat <base>..HEAD so you know exactly what is going to the …

    line 52
  6. Distinctive:

    Delete short-lived branches yourself once they are merged everywhere they need to go.

    line 71

“Common” is measured: kept by at least 6 in 10 of the 6 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

Gitflow

This skill covers the full gitflow branching model with two long-lived branches and short-lived branches off them.

  • main is production. Merging to main triggers your deploy pipeline. Treat every merge to main as a production deploy that your users will see shortly after.
  • dev (sometimes called develop) is the integration branch. Day-to-day feature and fix work lands here first.
  • Short-lived branches: feat/<desc> and fix/<desc> off dev, release/<x.y.z> off dev, hotfix/<x.y.z> off main.

Pick the flow

Read the matching reference in full and follow it step by step. Each one is self-contained.

The user wants to...FlowRead
Ship the accumulated dev work to productionReleasereferences/release.md
Get an urgent fix into production now, can't wait for a releaseHotfixreferences/hotfix.md
Build a new feature or a fix that will ride the next releaseFeaturereferences/feature.md

If the intent is ambiguous (e.g. "ship the settings fix"), ask one question: is this urgent enough to go straight to production (hotfix), or does it ride the next release (feature now, release later)? Then stop and wait for the answer. Don't pick a flow, name a branch, run a command or propose a version until the user replies.

Universal rails

These apply to every flow. They exist because main is live production and a mistake here is user-visible, so the cost of a shortcut is real.

  • Requires gh (brew install gh) for PRs, merges, and releases. Confirm gh auth status succeeds before the first PR; if not, stop and tell the user to run gh auth login.
  • main is a deploy button. Never merge to main while any CI check is not green. The full Definition-of-Done gate runs on a PR to main; wait for all of it. Poll CI every few minutes rather than tight-looping, to keep token cost sane. No CI pipeline configured on the repo? Skip the wait-for-green step and say so explicitly before merging.
  • Always update CHANGELOG.md during a release and a hotfix (never per-feature; the changelog is assembled at ship time). See references/changelog.md for the format.
  • Conventional Commits. Follow the Conventional Commits spec. Keep commit messages, PR titles, PR bodies, the changelog, and release notes clean and professional.
  • Never force-push. Never bypass git hooks (--no-verify or equivalent). The pre-push hooks are part of the safety net. If a push fails due to a hook, investigate and root-fix the underlying issue; never bypass.
  • Stage only the files you intend. Prefer git add <specific-files> over blanket staging commands. Verify what you are about to stage with git diff --staged before committing.
  • Verify before every push. Run git log --oneline <base>..HEAD and git diff --stat <base>..HEAD so you know exactly what is going to the remote, especially before anything that touches main.
  • After any merge to main (release or hotfix), open a break-glass revert PR and leave it OPEN. It is the one-click rollback. Never merge it yourself. See the shared procedure in references/release.md (the revert step is identical for hotfix).

Versioning

Semantic versioning. The current version lives in your project manifest (package.json, pyproject.toml, Cargo.toml, etc.) on both long-lived branches (kept in sync; both are at the last released version). No version manifest in the repo? Version via the latest git tag alone and skip the bump-file step.

  • Release: bump minor for new features (e.g. 1.1.0 to 1.2.0), major for breaking changes.
  • Hotfix: bump patch (e.g. 1.1.0 to 1.1.1).
  • The tag is v<x.y.z>, annotated, placed on the merge commit on main.

Merge methods (they differ by target, on purpose)

MergeMethod
Feature/fix PR into devSquash merge (one clean commit per PR with the (#NNN) suffix). This keeps dev history readable.
release/* or hotfix/* into mainMerge commit (--no-ff). This preserves the release as a first-parent merge so the break-glass revert (git revert -m 1) is clean.
Back-merge release/* or hotfix/* into devMerge commit.

Delete short-lived branches yourself once they are merged everywhere they need to go.

Stay in the loop

These flows touch production. When CI is running or a deploy is in flight, report each milestone to the user as it lands (CI green, merged, deploy healthy, tagged) rather than going silent. If any gate is red, stop and root-fix; do not merge around it.

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.