From agent-harness-skills
Splits changes into atomic commits with scoped validation, task-state updates, and Conventional Commit messages. Prevents unrelated files and ensures minimal, reviewable commits.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agent-harness-skills:atomic-commit-disciplineThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Turn completed work into small, reviewable commits with scoped validation, related task-state updates, and no unrelated files.
Turn completed work into small, reviewable commits with scoped validation, related task-state updates, and no unrelated files.
This skill is not a review template or ledger template. It focuses on minimal commits, exact staged paths, task-state coupling, validation, and commit messages. For shared harness terms, see ../../references/harness-patterns.md; when commit rules are absent, use references/build-when-missing.md. For delivery artifact evidence, see references/delivery-artifact-checklist.md.
git status or diffs to avoid unrelated work.git status, git diff, and git diff --cached.tasks.md, docs/tasks.md, or the repository's issue tracker reference.git status and relevant diffs to identify your changes and unrelated changes.git status --short --branch and list all tracked and untracked changes.git diff -- <path> for every file you intend to commit; do not commit changes you do not understand.references/build-when-missing.md.git add <exact paths>; avoid git add . unless the whole worktree contains only this change.git show --stat --oneline HEAD and git status --short.agent_chats in the same commit, it is staged with the change.# Atomic Commit Plan
## Detected Mapping
- work-state:
- ledger:
- validation:
## Worktree State
-
## Commit Groups
1. Subject:
Paths:
Task-state updates:
Validation:
## Excluded Changes
-
## Final Checks
- First:
- Then:
- Finally:
git add . and bringing unrelated files into the commit.Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub yfge/codex-plugins --plugin agent-harness-skills