Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/dgilford/ai-science-toolkit/commit-batchingnpx skills add dgilford/ai-science-toolkit --skill commit-batchinggit clone --depth 1 https://github.com/dgilford/ai-science-toolkitWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00072 | $0.01208 |
| Opus 5 | $0.00036 | $0.00604 |
| Sonnet 5 | $0.00014 | $0.00242 |
| Haiku 4.5 | $0.00007 | $0.00121 |
Grade A, and why
commit-batching scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are about to commit. Do not git add -A && git commit the whole tree at once. Group the pending changes into logical, single-concern commits first.
Current working tree
git status --short 2>/dev/null || echo "(not a git repo)"
echo "--- branch ---"; git rev-parse --abbrev-ref HEAD 2>/dev/null || true
echo "--- staged/unstaged stat ---"; git diff --stat HEAD 2>/dev/null || true
echo "--- untracked ---"; git ls-files --others --exclude-standard 2>/dev/null || true
Method
- Survey — run
git status(staged, unstaged, and untracked) and read enough ofgit diffto understand each change's intent. Never stage what you haven't looked at. - Plan — group changed paths into single-concern batches and draft a one-line message for each. Present the plan as a compact list before executing.
- Execute per group —
git add <explicit paths>→ confirmgit diff --cached --statmatches the intended group → commit with a focused message. Repeat for the next group. - Verify tree — after the loop,
git statusshould be clean (or leave only paths you deliberately chose not to commit). - Push — only if the original request asked you to push. Respect the repo's branching convention (branch first unless the project says commit straight to main).
Grouping principles
- One concern per commit. A feature, a bugfix, a refactor, and a config/dependency bump are separate commits even when touched in the same working session.
- Documentation folds in. A README or doc update that documents a specific batch belongs inside that batch's commit, not a standalone "update docs" commit. If one doc file's edits span several batches and can't be cleanly split, fold the whole doc change into the last batch you commit.
- Generated/mechanical changes (lockfiles, formatting sweeps, generated tables) go either with the change that caused them or in their own clearly-labeled commit — never silently folded into a feature commit.
- Order for coherence. Put prerequisites (a helper, a rename, a refactor) before the change that consumes them, so each commit stands on its own.
- Don't over-split. One logical change spread across N files is one commit, not N.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 63 lines · 72 tokens per session scan A 28d8e93ca3ce
commit-batching is a skill published in the GitHub repository dgilford/ai-science-toolkit (62 stars, last pushed 13d ago), licensed MIT. It adds 72 tokens to every session and 1,208 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
git-wrapup
Land working-tree changes as logical commits — the work grouped by concern, topped by a release commit (version bump, changelog, regenerated artifacts) and an annotated tag. Verify, commit, tag. Stops at "committed and tagged locally" — no push, no publish. The release-and-publish skill picks up from here. Distilled…
wrap
Session end - document updates, commit.
ship
Commit, push, create PR, then after approval check CI and merge. Use when work is done and ready to ship.
campaign-commit
Choose the correct commit command in a campaign workspace. Use when you are about to commit and need to select camp commit, camp p commit, fest commit, or intentional root pointer sync via camp refs-sync.
close
The "wrap" verb of the loop — end the session by writing the journal entry, updating touched project READMEs, committing, and pushing. Auto-invoked as the tail of /spec and /implement (the producer verbs) on clean completion, so "wrap" is not a verb the user types; also runnable directly. Dispatches /implement-audit…
git-commit-helper
Generates meaningful git commit messages for Spec-Driven Development workflows. Use when creating commits, suggesting commit messages, or helping with task-based version control. Automatically links commits to work packages.