| name | commit |
| description | Create actual Waitkit Git commits by inspecting changed files, splitting logical units, staging the right files, and running git commit with project Conventional Commit messages. Use only when the user explicitly asks to commit. |
| allowed-tools | Bash(git *:*), Bash(bash *discover-changed-areas.sh:*), Read |
Waitkit Commit
Use this skill only when the user explicitly asks to create commits. For message
recommendations without committing, inspect git state and answer directly
without running git add or git commit.
Step 1: Inspect Git State
git status --short
git diff --stat
git diff --staged --stat
If the user asks to commit staged changes only, use git diff --staged as the
source of truth.
Step 2: Determine Scope and Type
Run changed-area discovery:
bash "${CLAUDE_SKILL_DIR}/scripts/discover-changed-areas.sh" auto
Read ${CLAUDE_SKILL_DIR}/references/scope-guide.md and choose the most
specific useful scope. Use chore(harness) for .claude/** harness changes.
Step 3: Split Logical Units
Split commits when changes are independent:
- Core package behavior or tests
- Documentation-only changes
- Changesets, changelog, or release metadata
- Tooling/configuration changes
- Harness changes
Do not split files that are part of one requested change.
Step 4: Stage and Commit
For each logical unit:
- Stage only relevant files with
git add.
- Commit with
git commit -m "<type>(<scope>): <description>".
- Verify with
git log --oneline -n 3.
Rules
- Do not run
git push.
- Do not commit generated caches or
dist/ output unless explicitly requested.
- If there are unrelated user changes, leave them unstaged and report them.