How to use the Git Flow Branch Creator agent skill

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent can load when a request matches the skill. This one is named git-flow-branch-creator and is described as an intelligent Git Flow branch creator that analyzes git status and diff and creates appropriate branches following the nvie Git Flow branching model.

When you ask the agent to start work on your uncommitted changes, the skill tells it to run git status to see the changed files, then git diff for unstaged changes or git diff --cached for staged changes. It classifies the work, generates a semantic branch name, creates the branch, switches to it, and summarizes its reasoning and next steps.

How does the skill pick a branch type?

The skill uses a short decision process. If the changes look like critical fixes for production issues, it considers a hotfix branch. If they look like release preparation such as version bumps or final tweaks, it considers a release branch. Otherwise it defaults to a feature branch.

The source describes these branch types and their base branches. Feature branches start from develop and merge back to develop. Release branches start from develop and merge to both develop and master. Hotfix branches start from master and merge to both develop and master.

• Feature: new functionality, UI/UX improvements, new API endpoints, non-breaking schema additions, new configuration options, non-critical performance work
• Release: version number changes, build configuration updates, documentation finalization, minor fixes before release, release notes, dependency version locks
• Hotfix: security vulnerability fixes, critical production bugs, data corruption fixes, service outage resolution, emergency configuration changes

Install the skill from the ZIP or the Claude apps

Download the ZIP file for the skill from this page. Unzip it so that the skill folder sits inside one of these locations: ~/.claude/skills/ to make it available across your projects, or .claude/skills/ inside a specific project to make it available only there.

If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the skill there.

• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into .claude/skills/ in the project root
• Claude apps: Settings, then Capabilities, then Skills, then upload

Check what you need before you run it

The skill needs git installed and available on the command line, and it must run inside a git repository that has changes to analyze. Without changes there is nothing to classify. The skill's own edge-case rules say that if no changes are detected, the agent should tell you and suggest checking git status or making changes first.

The skill does not run scripts and does not use the network, according to this listing. Its only actions are the git commands described in its workflow.

• git installed and on the command line
• A git repository with changes to analyze

How do you trigger the skill with a request?

You do not need a special command. Describe what you want in plain language while your changes are sitting in the working tree or staging area. The agent should recognise the request as matching the skill's purpose of starting a branch for your changes.

Good requests say what you want to happen. Some examples follow. The exact wording is up to you, and the agent decides whether the skill applies.

• Start a branch for my changes.
• I forgot to make a branch before editing. Can you create the right Git Flow branch for what I changed?
• Look at my staged changes and create a branch for them.
• Create a Git Flow branch for these changes and tell me why you chose that type.

What does a typical run look like?

Suppose you added a new user registration API endpoint. The skill's worked example classifies this as new functionality and additive work that is not critical, so it chooses a feature branch named feature/user-registration-api created from develop. Its command for that case is shown below.

Other examples in the skill cover a critical authentication security fix, which becomes hotfix/auth-security-patch from master, and a version bump with finalized release notes, which becomes release-2.1.0 from develop. Read the summary the agent gives you before you start committing, because the choice is a judgement, not a guarantee.

What safety checks should you apply?

The classification is heuristic, so it can be wrong. The listing warns against using this skill for hotfix decisions on production systems without human review, because a wrong branch type can cause a bad release path. Review the proposed type and name before accepting it, especially for anything that touches production.

The skill also tells the agent to handle mixed changes by prioritizing the most significant change type or suggesting a split into several branches, and to handle an existing branch by asking whether a new branch is needed. If a suggested name already exists, it should append an incremental suffix or propose another name. Check the result with git status and git branch after the agent finishes.

• Do not use it on projects that do not follow Git Flow, such as trunk-based or GitHub Flow
• Review hotfix classifications on production code by a person
• Split mixed changes yourself if the agent suggests it