git-commit

Sync with the tracked upstream, stage everything, read the diff, write a good commit message, and commit.

Steps

1. Inspect state and sync: git status --short --branch, then resolve the tracked upstream with git rev-parse --abbrev-ref --symbolic-full-name "@{upstream}".
   • If an upstream exists, run git pull --rebase --autostash.
   • If the pull, rebase, or autostash fails or leaves unmerged paths (git diff --name-only --diff-filter=U), stop without staging/committing and report the recovery state.
   • If no upstream is configured, skip the pull and say so.
2. Verify there are changes after pulling: git status --porcelain. If empty, stop and say so.
3. Stage all changes: git add -A.
4. Read the staged diff for context: git diff --cached --stat and git diff --cached.
5. Write the message:
   • Subject: concise, imperative, ≤50 chars. Conventional Commit prefix optional.
   • Body (when the change is non-trivial): blank line, then wrapped lines explaining what and why, not how.

6. Commit with a real multi-line message:

   git commit -m "$(cat <<'EOF'
   <subject>
   
   <body line>
   <body line>
   EOF
   )"

7. Confirm with git log -1 --stat and git status --short --branch.

Rules

• Plain text only — no markdown, code fences, or quotes inside the message.
• Group unrelated changes into one coherent message; don't list every file.
• Don't add Co-Authored-By or tool attribution unless asked.
• Never bypass a failed pull or unresolved conflict to create the commit.
• Never git push — committing only.