agentsclimarketplace

Commit

Skill ph3on1x/wisci/skills/commit

Creates conventional git commits enriched with a Context section that logs AI-layer changes (.wisci/, CLAUDE.md, skills, rules), turning git history into queryable long-term memory. Use to commit changes, optionally pushing afterward.From its SKILL.md

Install
npx -y skills add ph3on1x/wisci --skill commit

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

3 things to look at

  • reads credentialsReads from 5 credential sources: `.env` and 4 more.
  • 1 stars1 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.
  • runs commandsInstructs the agent to run 8 commands, including `git status` and 7 more.

SKILL.md

4.8 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

/commit — Context-Enriched Committer

Create git commits that track AI context evolution alongside application changes. When AI-layer files are part of the commit, a Context: section is appended to the commit body — making git log long-term, queryable memory for future sessions and agents.

Current state

!git status

!git --no-pager diff --stat HEAD

Recent commits (style reference): !git log --oneline -5

If the blocks above show errors or unexpanded variables, run the commands yourself with Bash.

Execution Flow

  1. Parse arguments. If $ARGUMENTS starts with the standalone word push (case-insensitive): enable push-after-commit, use the remainder as the commit message (may be empty).

  2. Check for committable changes in the state above. Nothing staged, unstaged, or untracked → report "Nothing to commit — working tree clean." and stop. Run git --no-pager diff HEAD for the full diff when the stat summary is not enough to write an accurate message.

  3. Classify every changed and untracked file:

    • Application code — stage normally
    • AI context files (paths below) — stage normally; flag for the Context section
    • Sensitive files (.env, *.key, *.pem, credentials.*, *secret*) — do not stage; warn the user
    • Unrelated concerns — if changes span multiple unrelated concerns and $ARGUMENTS gives no hint, ask before bundling
  4. Stage files individually (git add <file>). Never git add -A or git add ..

  5. Detect AI context changes. Check git --no-pager diff --cached --name-only against the AI Context Paths below.

  6. Draft the message.

    • $ARGUMENTS provided: use it; prepend a conventional prefix (feat:, fix:, refactor:, docs:, test:, chore:) if missing, inferred from the changes.
    • Empty: auto-generate from the staged diff — focus on why, not what.
  7. Append the Context section when AI-layer files are staged (omit entirely otherwise):

    Context:
    - <create|update|remove|rename> <path> — <what changed>
    
  8. Commit via heredoc:

    git commit -m "$(cat <<'EOF'
    <prefix>: <description>
    
    <optional body>
    
    Context:
    - <action> <path> — <annotation>
    EOF
    )"
    
  9. Verify and report: short hash, subject line, whether a Context section was included and how many AI-layer files it tracked.

  10. Push (if requested). git push; no upstream → git push -u origin <current-branch>. On failure: report the error, never retry with --force.

AI Context Paths

  • .wisci/** — WISCI context store (context files, handoffs, primer, index)
  • CLAUDE.md, .claude/CLAUDE.md — project instructions
  • .claude/rules/**, .claude/commands/**, .claude/skills/**, .claude/agents/**, .claude/hooks/** — Claude project config
  • .claude/settings.json, .claude/plugins/** — Claude settings and plugins
  • .mcp.json — MCP server config
  • skills/**, .claude-plugin/**, hooks/** — plugin/skill definitions (when the repo is itself a plugin)
  • .agents/** — cross-client agent skills
  • .cursorrules, .cursor/rules/** — Cursor rules
  • .github/copilot-instructions.md — Copilot instructions
  • AGENTS.md, codex.md, .codex/** — Codex instructions
  • GEMINI.md, .gemini/** — Gemini CLI instructions and config

A staged file matches if its path starts with a directory prefix above or exactly matches a filename entry.

Examples

Application + context changes:

feat: add token refresh to auth middleware

Automatic refresh when access token expires. One retry before 401.

Context:
- update .wisci/context/auth-research.md — added token refresh decision and rationale

Context-only changes:

docs: persist Stripe integration research

Context:
- create .wisci/context/stripe-integration.md — /isolate research results
- update .wisci/handoff/payments.md — Stripe added to in-progress work

Key Constraints

  • Never --no-verify. Pre-commit hook fails → report and stop.
  • Never force-push.
  • Never git add -A / git add .. Stage by name.
  • Never stage sensitive files.
  • Context section is conditional — only when AI-layer files are actually staged.
  • One logical commit — ask before bundling unrelated concerns.
  • Respect user intent — a provided message is used as-is apart from prefix and Context additions.
  • All diffs with --no-pager.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 325,949. 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.