git-commits

A set of rules saying that Git commits are made by a person, not by the AI. The AI may inspect changes and prepare commands or a commit message, but it does not stage, commit, amend, or push them by default.

In plain words
What is it for?
Use it whenever changes are ready to record in Git. It supports checking status and diffs, drafting a commit message, listing files to stage, and warning about sensitive files.
Why use it?
It prevents accidental changes to project history and gives the human final control over what is recorded and shared. It also calls attention to secrets that should not enter a commit.

Cursor rule for Cursor

Install

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.

agentmods
npx agentmods add rules/arxdsilva/vault/git-commits
Clone the repo
git clone --depth 1 https://github.com/arxdsilva/vault

Made for: Cursor.

Per session 1,004 This file is loaded in full into every session.
When invoked 1,004 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.01004 $0.01004
Opus 5 $0.00502 $0.00502
Sonnet 5 $0.00201 $0.00201
Haiku 4.5 $0.00100 $0.00100

Measured yesterday against content hash 88bc9c37d0cf, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

git-commits 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 yesterday.

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.

.cursor/rules/git-commits.mdc · 112 lines

How it starts

The opening of the file, as written. The whole thing — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Git Commit Practices

Commits are a human action. The AI must never run git commit, git add (to stage for a commit), or git commit --amend — even when the user says "commit this" or "commit your changes."

When commits are in scope, the AI may inspect the repo and prepare a commit for the human; the human runs the final git commands locally.

What the AI Must Never Do

  • Run git commit or git commit --amend
  • Run git add to stage files for a commit
  • Create commits to "save progress" or after finishing a task
  • Create empty commits
  • Push to the remote unless the user explicitly asks (see Push below)

If the user asks you to commit, decline the git command and instead offer a ready-to-run suggestion (message + git add / git commit commands they can copy).

What the AI May Do

  • Run read-only inspection: git status, git diff, git log
  • Draft a commit message and list which paths to stage
  • Warn about secrets (.env, credentials.json, tokens, keys) that must not be committed
  • Fix hook/lint issues in code when asked — the human commits after validation

Git Safety Protocol

When any git write command is in scope (e.g. push, branch ops), still follow:

  • Never update git config (git config, global or local)
  • Never run destructive or irreversible commands (push --force, reset --hard, etc.) unless the user explicitly requests them
  • Never skip hooks (--no-verify, --no-gpg-sign, etc.) unless the user explicitly requests it
  • Never force-push to main or master; warn the user if they ask
  • Never use interactive git (git add -i, git rebase -i)

Suggesting a Commit (human runs these)

When the user wants to commit, or work is ready to land, follow this sequence.

1. Inspect (run in parallel)

git status
git diff          # staged and unstaged
git log -10 --oneline   # match recent message style

Do not run extra exploration commands beyond these git commands.

2. Draft the message

Read the full file on GitHub · 112 lines

Changes

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.

  1. yesterday First seen · 112 lines · 1,004 tokens per session scan A 88bc9c37d0cf

Subscribe to this mod's changes

git-commits is a cursor rule published in the GitHub repository arxdsilva/vault (3 stars, last pushed 1mo ago), licensed MIT. It adds 1,004 tokens to every session, about $0.0050 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 2026-08-31.