commit Skill
Staged Änderungen und generiert formatierte Commit-Messages gemäß Repository-Konventionen und Pre-commit Hooks.
- GitHub Stars
- 30
- Dateien
- 1
- Letzter Commit
- 3. Nov. 2026
- Lizenz
- MIT
- Erlaubte Tools
- 13
Was macht der commit Skill?
commit ist ein Agent Skill. Staged Änderungen und generiert formatierte Commit-Messages gemäß Repository-Konventionen und Pre-commit Hooks.
Zusammenfassung erzeugt aus der SKILL.md des Skills.
Über den commit skill
Entwickler, die an Pull Requests arbeiten, können diesen Skill nutzen, um das Staging und Committen von Code-Änderungen zu automatisieren und dabei spezifische Repository-Standards einzuhalten. Das Tool verwaltet den gesamten Git-Workflow, indem es Konfigurationsdateien wie .git-workflow/config.yaml oder .claude/config.yaml ausliest, um erforderliche Commit-Formate, erlaubte Change-Typen und Ticket-Anforderungen zu bestimmen. Es nutzt git status, diff und add Befehle, um Dateien vorzubereiten, und kann automatisch versuchen, Fehler von Pre-commit Hooks zu beheben, indem es Linting- oder Formatting-Tools wie ESLint, Prettier, Black oder Ruff ausführt.
Wenn ein Agent diesen Skill lädt, agiert er als spezialisierter Git-Assistent, der notwendigen Kontext wie Ticket-IDs und Change-Typen durch Interaktion mit dem Nutzer sammelt. Er erstellt korrekt formatierte Commit-Messages, wie zum Beispiel [Fix] message (PROJ-1234), und pflegt eine lokale pr-context.json Datei, um die Commit-Historie und Branch-Informationen zu tracken. Dies stellt sicher, dass jeder Commit konsistent mit den Projektkonventionen bleibt, ohne dass manuelle Formatierungsfehler entstehen.
Was ein agent damit machen kann
- 1Änderungen stagen und committenAlle geänderten Dateien stagen und eine formatierte Commit-Message basierend auf den Projektregeln erstellen.
- 2Pull Request Kontext beibehaltenEine lokale JSON-Datei mit Commit-Hashes und Zeitstempeln aktualisieren, um den Fortschritt auf einem Branch zu tracken.
Geschrieben basierend auf dem SKILL.md des skills. Jeder use case wurde gegen ein zitat daraus geprüft.
So installierst du den commit Skill
Jeder Befehl kopiert den Skill-Ordner in das Skills-Verzeichnis des Clients. Starte den Client neu oder beginne eine neue Session, um ihn zu laden.Den commit Skill in Plugin installieren
/plugin marketplace add rlajous/claude-code-commands /plugin install git-workflow@git-workflow-marketplace
Den commit Skill in Claude Code installieren
~/.claude/skills/commit/SKILL.md
$ npx degit rlajous/claude-code-commands/skills/commit ~/.claude/skills/commitDen commit Skill in Claude Code (project) installieren
.claude/skills/commit/SKILL.md
$ npx degit rlajous/claude-code-commands/skills/commit .claude/skills/commitDen commit Skill in Codex installieren
~/.agents/skills/commit/SKILL.md
$ npx degit rlajous/claude-code-commands/skills/commit ~/.agents/skills/commitDen commit Skill in Gemini CLI installieren
~/.gemini/skills/commit/SKILL.md
$ npx degit rlajous/claude-code-commands/skills/commit ~/.gemini/skills/commitDen commit Skill in Claude.ai installieren
npx degit rlajous/claude-code-commands/skills/commit commit && zip -r commit.zip commit # claude.ai → Customize → Skills → + → Create skill → Upload a skill → commit.zip
SKILL.md
Auf GitHub ansehenCross-runtime: follow runtime compatibility for invocation, delegation, configuration precedence, state paths, and permissions.
You are helping commit changes for a pull request. Your task is to stage changes, generate a properly formatted commit message following project conventions, and handle pre-commit hooks.
Step 1: Load Configuration
Check for configuration and context:
# Check for config and context files
if [ -f ".git-workflow/config.yaml" ]; then CONFIG_PATH=".git-workflow/config.yaml"; elif [ -f ".claude/config.yaml" ]; then CONFIG_PATH=".claude/config.yaml"; else CONFIG_PATH=""; fi
if [ -f ".git-workflow/pr-context.json" ]; then CONTEXT_PATH=".git-workflow/pr-context.json"; elif [ -f ".claude/.pr-context.json" ]; then CONTEXT_PATH=".claude/.pr-context.json"; else CONTEXT_PATH=""; fi
Load from the resolved CONFIG_PATH (canonical first, legacy read-only fallback):
commits:
format: "[{type}] {message} ({ticket})"
types: [Feature, Fix, Hotfix, Refactor, Docs, Test, Chore]
requireTicket: false
ticketPattern: "^[A-Z]+-\\d+$"
attribution:
enabled: false
format: ""
Load from the resolved CONTEXT_PATH (canonical first, legacy read-only fallback):
{
"ticket_id": "PROJ-1234",
"type": "fix",
"branch": "fix/proj-1234-description"
}
Default Values (when no config):
commits:
format: "[{type}] {message} ({ticket})"
types: [Feature, Fix, Hotfix, Refactor, Docs, Test, Chore]
requireTicket: false
attribution:
enabled: false
Step 2: Gather Context
If .pr-context.json doesn't exist or is incomplete, ask for missing information:
Missing Ticket ID
If commits.requireTicket: true or no ticket found in context:
Question: "What is the ticket ID for this commit?"
- Example:
PROJ-1234,ENG-456 - Allow skipping if
requireTicket: false
Missing Change Type
If type not found in context:
Question: "What type of change is this?"
Options: Feature, Fix, Hotfix, Refactor, Docs, Test, Chore
Step 3: Show Changed Files
Display what will be staged:
# Show git status
git status --short
# Show detailed diff stats
git diff --stat
git diff --cached --stat
Present to user:
Changes to be committed:
Staged:
M src/file1.ts
A src/file2.ts
Unstaged:
M src/file3.ts
M tests/file1.test.ts
Do you want to stage all changes, or select specific files?
Options:
- Stage all changes (default)
- Stage only unstaged changes
- Select specific files
- Review diff before staging
If user wants to review diff:
git diff
Step 4: Stage Files
Based on user's choice:
Stage All
git add -A
Stage Specific Files
git add {file1} {file2} ...
Confirmation
# Show what's staged
git diff --cached --name-only
Step 5: Generate Commit Message
Ask for Summary
Question: "What is the summary of your changes?"
Guidelines:
- Use imperative mood ("Add feature" not "Added feature")
- Be concise (50-72 characters)
- Focus on what and why, not how
Examples:
- "Add user authentication with OAuth2"
- "Fix memory leak in connection pool"
- "Update dependencies to latest versions"
Format Message
Apply commit format from config (default: [{type}] {message} ({ticket})):
| Variable | Source | Example |
|---|---|---|
{type} |
Context or user input | Fix |
{message} |
User-provided summary | Fix login timeout |
{ticket} |
Context or user input | PROJ-1234 |
Type Capitalization:
fix->Fixfeature->Featurehotfix->Hotfix
Result Examples:
[Fix] Standardize token_info schema (PROJ-1234)[Feature] Add holder analysis endpoint (ENG-456)[Chore] Update dependencies(no ticket)
Handle Missing Ticket
If no ticket and format includes ({ticket}):
- If
requireTicket: true: Prompt for ticket - If
requireTicket: false: Omit the({ticket})part
Without ticket: [Fix] Update dependencies
With ticket: [Fix] Update dependencies (PROJ-1234)
Step 6: Commit Changes
Create Commit
git commit -m "{formatted_commit_message}"
Important: Let pre-commit hooks run. Do NOT use --no-verify.
Handle Pre-commit Failures
If commit fails due to pre-commit hooks:
-
Identify the issue:
# Check hook output for errors # Common issues: linting, formatting, type errors -
Auto-fix if possible:
# For ESLint/Prettier npm run lint:fix 2>/dev/null || pnpm lint:fix 2>/dev/null || yarn lint:fix 2>/dev/null # For Python black . 2>/dev/null || ruff format . 2>/dev/null -
Re-stage fixed files:
git add -A -
Retry commit:
git commit -m "{formatted_commit_message}" -
If still failing:
- Show the error to the user
- Ask if they want to fix manually or skip hooks (not recommended)
Attribution (If Enabled)
If attribution.enabled: true in config, append the configured prose note to the commit body. The note must comply with repository commit rules and must not impersonate or add another commit author.
git commit -m "{message}
{attribution.format}"
Example:
[Fix] Update token schema (PROJ-1234)
This commit was created with assistance from an AI coding agent.
Note: Attribution is disabled by default.
Step 7: Update Context
Update .git-workflow/pr-context.json:
mkdir -p .git-workflow
{
"ticket_id": "PROJ-1234",
"ticket_url": "https://...",
"branch": "fix/proj-1234-description",
"type": "fix",
"description": "Short description",
"started_at": "2025-01-17T12:00:00Z",
"commits": [
{
"hash": "abc123",
"message": "[Fix] Summary (PROJ-1234)",
"timestamp": "2025-01-17T12:30:00Z"
}
],
"last_commit_at": "2025-01-17T12:30:00Z"
}
Step 8: Confirm
Output a confirmation message:
Staged: {N} files
Committed: {commit_hash_short}
Message: {commit_message}
Next step: Run /finish to create the pull request
If there were pre-commit fixes:
Pre-commit hooks auto-fixed some issues:
- Formatted 3 files with Prettier
- Fixed 2 ESLint warnings
Staged: 5 files
Committed: abc1234
Message: [Fix] Update token schema (PROJ-1234)
Configuration Reference
| Setting | Default | Description |
|---|---|---|
commits.format |
[{type}] {message} ({ticket}) |
Commit message format |
commits.types |
Feature, Fix, Hotfix... | Allowed commit types |
commits.requireTicket |
false |
Require ticket ID |
attribution.enabled |
false |
Add an AI assistance note |
attribution.format |
empty | Optional prose note |
Error Handling
| Scenario | Action |
|---|---|
| No changes to commit | Inform user, suggest checking git status |
| Pre-commit hook fails | Attempt auto-fix, show errors if still failing |
| Ticket required but missing | Prompt for ticket ID |
| Invalid commit type | Show allowed types, ask again |
| Context file missing | Proceed with user input, create new context |
| Git not in repo | Error with clear message |
Common Commit Formats
The command supports various commit format patterns:
Conventional Commits
commits:
format: "{type}: {message}"
# Result: fix: update token schema
With Scope
commits:
format: "{type}({scope}): {message}"
# Result: fix(api): update token schema
Ticket Prefix
commits:
format: "{ticket}: {message}"
# Result: PROJ-1234: Update token schema
Custom Format
commits:
format: "[{ticket}] {type} - {message}"
# Result: [PROJ-1234] Fix - Update token schema
Examples
Standard Flow
User: /commit
Agent: [Shows staged/unstaged files]
Agent: Stage all changes? -> Yes
Agent: What is the summary? -> "Fix login timeout issue"
Result: Committed abc1234
Message: [Fix] Fix login timeout issue (PROJ-1234)
With Auto-fix
User: /commit
Agent: [Stages files]
Agent: Summary? -> "Update API response"
[Pre-commit fails - ESLint errors]
Agent: Auto-fixing lint errors...
Agent: Fixed 3 files, retrying commit...
Result: Committed def5678
Message: [Fix] Update API response (PROJ-1234)
No Context File
User: /commit
Agent: No context found. What's the ticket ID? -> "ENG-789"
Agent: What type of change? -> "Feature"
Agent: Summary? -> "Add dark mode support"
Result: Committed ghi9012
Message: [Feature] Add dark mode support (ENG-789)
Dateien in diesem Skill
1commit Skill-Fragen
Wie installiere ich den commit Skill in Claude Code?
Führe npx degit rlajous/claude-code-commands/skills/commit ~/.claude/skills/commit aus. Claude Code lädt Skills aus ~/.claude/skills (alle Projekte) oder .claude/skills (ein Projekt) und nutzt einen Skill, wenn seine Beschreibung zur Aufgabe passt.
Welche Clients können den commit Skill nutzen?
Jeder Client, der das Agent Skills-Format liest: Claude Code, Codex, Gemini CLI und claude.ai (als ZIP hochladen). Skripte im Skill laufen nur dort, wo der Client Codeausführung erlaubt.
Kann ich das Format der Commit-Messages anpassen?
Ja, du kannst das Commit-Format, erlaubte Typen, Ticket-Anforderungen und Attribution-Einstellungen in einer Konfigurationsdatei festlegen.
Unterstützt dieser Skill die Kennzeichnung von KI-Unterstützung?
Ja, wenn es in der Konfiguration aktiviert ist, kann ein Textbaustein im Commit-Body angehängt werden, der darauf hinweist, dass der Commit mit KI-Unterstützung erstellt wurde.