productivity-kit: Skill for Claude Code

Commit

Safely commit the current working changes with a concise, descriptive message.

Step 1 — Gather the diff

git status --short
git diff HEAD

If there are no changes (git status reports nothing), stop and tell the user "Nothing to commit — working tree is clean."

Step 2 — Scan for sensitive data

Before doing anything else, scan the full diff for the patterns below. If ANY match is found, stop immediately, show the user the exact file and line, and do NOT commit.

Patterns to reject:

Category	Examples to detect
.env file staged	diff --git a/.env or diff --git a/.env.
API keys	tokens matching sk-[A-Za-z0-9]{20,}, AIza[0-9A-Za-z\-_]{35}, AKIA[0-9A-Z]{16}
GitHub PATs	ghp_[A-Za-z0-9]{36}, github_pat_
Generic secrets	variable names containing SECRET, PASSWORD, PASSWD, TOKEN, PRIVATE_KEY assigned to a string literal
Private keys	`-----BEGIN (RSA
Patch files staged	any file ending in .patch

Use Grep on the diff output for each pattern. If clean, proceed.

Step 3 — Stage changes

If nothing is staged yet (git diff --cached is empty), stage all tracked changes:

git add -u

Do not git add . — do not stage untracked files the user has not explicitly added. If the user wants to stage specific files, ask them which ones before staging.

Step 4 — Write the commit message

Analyse the staged diff and write a commit message that:

• Summarises the what and why in ≤ 72 characters on the first line.
• Uses imperative mood ("Add", "Fix", "Remove", "Update" — not "Added", "Adds").
• Optionally includes a blank line followed by a short body (2–4 bullet points) for complex changes — skip the body for small/obvious commits.
• Does not mention file names unless they are the entire point of the change.
• Does not start with "Commit", "Update files", or other filler phrases.

Step 5 — Commit

git commit -m "$(cat <<'EOF'
<first line — ≤72 chars>

- <optional bullet 1>
- <optional bullet 2>
EOF
)"