git-commit

Commit changes in the git repository

GOAL: use the git-add, git-message and git-status skills in sequence to commit changes in the repository and clean it.

WHEN: use when the user or agent needs to commit changes in the repository.

NOTE: Await user approval of commit messages before committing. Make commits only after the user has approved the commit message.

Workflow

Follow these steps in sequence:

Step 1: Stage atomic changes

• Invoke the git-add skill.
• Capture the status from skill output (SUCCESS, WARN, or ERROR).
• Handle the status:
  • If ERROR: Halt and report the error to the user.
  • If WARN (no files available): Skip to Step 5.
  • If SUCCESS (files staged): Continue to Step 2.

Step 2: Generate commit message and await approval

• Invoke the git-message skill.
• Capture the status from skill output (APPROVED, REJECTED_EDIT_FILES, REJECTED_REGENERATE, REJECTED_ABORT, or ERROR).
• Handle the status:
  • If ERROR: Halt and report the error to the user.
  • If APPROVED: Continue to Step 3.
  • If REJECTED_EDIT_FILES: Unstage all files with git restore --staged ., then loop back to Step 1.
  • If REJECTED_REGENERATE: Loop back to Step 2.
  • If REJECTED_ABORT: Halt and report abort to the user.

Step 3: Commit changes

IMPORTANT: To avoid a recursive loop, DON'T invoke the git-commit skill here.

• Without using the git-commit skill, perform a direct commit with git using the approved message.
• If success: report the commit SHA and continue to Step 4.
• If failure: Halt and report the error to the user.

Step 4: Check for more commits (automatic loop control)

• Run git status --porcelain=v2 to check for remaining changes.
• If no changes: Continue to Step 5.
• If changes remain: automatically loop back to Step 1 (no user prompt needed).

Step 5: Present final status

• Invoke the git-status skill.
• DONE

Efficiency directives

• Optimize all operations for agent, token, and context efficiency
• Batch operations on file groups, avoid individual file processing
• Use parallel execution when possible
• Target only relevant files
• Reduce token usage