メニュー
Safely commit changes. Reviews staged/modified files for sensitive data, writes a commit message with body. Never adds co-author lines.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Generate a PR title and description following the WCC repo template, infer change types from the diff, and print the GitHub URL to open the PR in the browser. Use when asked to open a PR or create a pull request.
Review pull requests using GitHub CLI. Use when asked to check a PR, perform code review, or leave review comments. Focus on bugs, regressions, security, code quality, best practices, and missing tests; post comments inline on the relevant changed lines and avoid top-level summary comments unless explicitly requested.
Review local staged and unstaged changes before committing. Outputs an overall summary and per-file line-level findings with severity levels. Use when asked to review code before a commit, check local changes, or suggest improvements on uncommitted work.
Create a GitHub issue in the Women-Coding-Community/wcc-frontend upstream repository, add it to the Frontend Platform project board, and set Epic, Status, and other project fields. Use when the user asks to create a GitHub issue, ticket, or task for this project.
Interactively create a well-structured GitHub issue (bug or feature) for Women-Coding-Community/wcc-frontend with Scrum user story, acceptance criteria, test scenarios, and code examples derived from the codebase. Adds the issue to the Frontend Platform project board.
| name | commit |
| description | Safely commit changes. Reviews staged/modified files for sensitive data, writes a commit message with body. Never adds co-author lines. |
| disable-model-invocation | true |
| allowed-tools | Bash |
Canonical runbook:
.ai/skills/commit.mdThis file is the Claude Code adapter. The workflow logic is defined in the canonical skill so it can be shared with other agents (Codex, Copilot, Cursor). Any changes to the workflow should be made in.ai/skills/commit.mdfirst.
You are now in commit mode. Follow these steps in order, stopping if any check fails.
Run these in parallel:
git status --shortgit diff HEADgit log --oneline -5If there are no staged or modified files, tell the user there is nothing to commit and stop.
Review ALL output from Step 1 carefully. Look for any of the following in file names or diff content:
Sensitive patterns to detect:
sk-, ghp_, xoxb-, AKIA, Bearer , token, api_key, apikey)password, passwd, secret, credential).pem, .key, .p12, .pfx).env, .env.local, .env.production, any .env*)aws_access_key, aws_secret, client_secret)If ANY sensitive data is detected:
Stop immediately and output:
WARNING: Sensitive data detected — commit aborted.
Found in: <file(s)>
Issue: <brief description of what was found>
Please remove the sensitive data before committing. Consider using environment variables or a secrets manager instead.
Do NOT proceed further.
Analyze the diff and status to write a commit message:
Format:
<subject line>
<body explaining why and context>
NEVER include any Co-Authored-By, Co-authored-by, or similar trailer lines.
Show the draft commit message to the user before proceeding.
Stage the specific files using git add <files> (never use git add . or git add -A — always add files by name to avoid accidentally including sensitive files).
Run:
git commit -m "$(cat <<'EOF'
<subject line>
<body>
EOF
)"
Do NOT append any Co-Authored-By or similar lines to the commit message.
After committing, run git status to confirm success.
git add . or git add -A