agentsclimarketplace

Commit

Skill neumie/almanac/skills/git/commit

Personal agent toolkit — curated skills for LLM coding agents

Install
npx -y skills add neumie/almanac --skill commit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when committing changes. Analyzes diff, writes a conventional commit message, commits immediately without confirmation. Triggers: commit, save my work, done with this change.

SKILL.md

3.0 KB, 659 tokens by cl100k_base, as published. Nobody here has run it

Commit

Commit staged and unstaged changes following the rules below.

Commit message format

Use semantic commit messages matching the project convention:

<type>(<scope>): <short summary in imperative mood>

Types: feat, fix, refactor, perf, chore, ci, docs, test

  • Summary is lowercase, no period at the end
  • Imperative mood ("add X", not "added X" or "adds X")
  • Keep the first line under 72 characters
  • Add a blank line + body only if the "why" isn't obvious from the summary

Splitting into logical commits

  • If you have context about what work was done (e.g. you just finished implementing something), commit only that work. Don't bundle unrelated changes — leave other unstaged/untracked changes alone.
  • Planning artifacts ARE part of the work — always commit them. A plan/spec/context scaffold that drove the change (e.g. docs/plans/<date>-<slug>/ containing context.md, README.md, .planning-prompt, or similar planning docs) is not "unrelated untracked junk" — it is the record of why this commit exists. Stage and commit these alongside the code in the same commit (or a dedicated docs: commit). Never leave them behind. The only exception is regenerated runtime logs explicitly listed in .gitignore.
  • You MAY split your work into multiple logical commits if it makes sense (e.g. a refactor commit + a feature commit, or separating test changes from implementation).
  • If you have no prior context, read git diff and git diff --cached to understand all changes, then group them into reasonable logical commits by topic/purpose. If some changes appear unrelated to each other and you can't determine what belongs together, ask before committing.
  • Each commit should be self-contained and buildable on its own when possible.

Procedure

These commands run automatically when the skill loads — output replaces each line below:

  • Working tree status: !git status
  • Unstaged diff: !git diff
  • Staged diff: !git diff --cached
  • Recent commits: !git log --oneline -10

Then:

  1. Review the changes and decide how to split them into commits (if needed).
  2. For each commit:
    • Stage the relevant files with git add <specific files> (never git add -A or git add .)
    • Create the commit
  3. Run git status after all commits to verify nothing was missed.

Edge cases

  • Nothing to commit: Report and stop.
  • Pre-commit hook failure: The commit did NOT happen. Fix the issue, re-stage, and create a new commit. Never use --amend after a hook failure — it would modify the previous commit.
  • Secrets / dangerous files: If you spot .env files, API keys, tokens, credentials, or private keys, stop and warn before committing.
  • Merge conflict markers: Warn and refuse to commit those files.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.