Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Agent Commit — measured AI agent skill | DecimalAI

majiayu000/agent-commit

NewSkillScoreMaintained

Analyze changes and create a meaningful commit with agent authorship. Internal skill for evolve_loop.

agent-toolingby majiayu000

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

DevOps & SRE pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/majiayu000-agent-commit/SKILL.md

decimalai skills pull majiayu000-agent-commit --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/majiayu000-agent-commit/SKILL.md https://app.decimal.ai/s/majiayu000-agent-commit/SKILL.md

Raw file: /s/majiayu000-agent-commit/SKILL.md ↗ · machine-readable summary at /s/majiayu000-agent-commit.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

23 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 2h ago

—+51% tokens0% turns61% of whole cases pass with skill (14/23)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-08	✗→✓	▲ Improved	118%	0%
case-10	✗→✓	▲ Improved	11%	0%
case-12	✗→✓	▲ Improved	57%	0%
case-14	✗→✓	▲ Improved	45%	0%
case-15	✗→✓	▲ Improved	100%	0%

Showing 5 of 23 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

majiayu000 / claude-skill-registry

skills/agent/agent-commit/SKILL.md

View on GitHub ↗

License MITImported 8/3/2026Version v1

Skill contents

SKILL.md

Agent Commit

Creates a git commit with meaningful message based on skill output and actual file changes.

When to Use

This skill is called automatically by evolve_loop.py after a content-modifying skill completes. It is not intended for manual invocation.

Input

You will receive:

The name of the skill that just ran
A summary or excerpt of that skill's output

Instructions

1. Check for Changes

Run git status --porcelain to see if there are uncommitted changes.

If there are no changes, output "No changes to commit" and stop.

2. Analyze the Changes

Run git diff (or git diff --cached if files are staged) to understand what was modified.

Focus on:

Which files were changed (especially content files in obsidian/)
The nature of the changes (new content, edits, deletions)
Key themes or topics affected

3. Generate Commit Message

Create a commit message with a summary line and body:

<type>(<skill>): <summary>

<body>

Summary line (required, under 72 chars):

<type> is one of: feat, fix, refine, research, chore
<skill> is the skill name (e.g., deep-review, condense, expand-topic)
<summary> describes WHAT was done and to WHICH file(s)
File paths must include parent directory (e.g., topics/free-will.md not just free-will.md)

Body (required for content changes):

2-5 bullet points summarizing key changes
For condense: before/after word counts, what was cut vs preserved
For deep-review: what issues were found and fixed
For expand-topic: key themes covered
For research: sources consulted, main findings

Good examples:

refine(condense): reduce topics/free-will.md from 9567 to 2985 words

- Cut quantum Zeno redundancy (7+ mentions → 1 via link)
- Removed extended sections on altered states, process philosophy
- Preserved core phenomenology arguments and Mary's Room
- Kept tenet connections and cross-links intact

refine(deep-review): improve clarity in concepts/qualia.md

- Fixed circular definition in opening paragraph
- Added concrete examples for inverted spectrum argument
- Strengthened connection to hard problem article

Bad examples:

auto(deep-review): Automated execution (too vague, no body)
Updated files (no context)
refine(condense): reduce free-will.md (missing parent directory - is it topics/ or concepts?)

4. Create the Commit

Stage all changes and commit with agent authorship. Use a HEREDOC for the multi-line message:

bash
git add -A
git commit --author "unfinishablemap.org Agent <agent@unfinishablemap.org>" -m "$(cat <<'EOF'
<summary line>

- bullet point 1
- bullet point 2
EOF
)"

5. Output the Result

Output the commit hash and message so the calling system can log it.

Important

Keep commit messages under 72 characters for the summary line
Always include parent directory in file paths (e.g., topics/free-will.md, concepts/qualia.md, research/notes.md)
Focus on the "what" and "which file" - the skill name already tells us "how"
If multiple files changed, mention the primary one and note others exist
Never include time estimates or subjective assessments in commit messages

Want to see if majiayu000/agent-commit would activate on your phrasing? Try it before you install — no LLM call, no signup.

Other measured skills in the registry, with their headline benchmark lift.