020-agent-audit

020-agent-audit is a cursor rule for Cursor from d-padmanabhan/agent-engineering-handbook. It costs 2,491 tokens per session, scanned A, original, MIT.

A set of rules for AI agents covering permission for remote changes, risk checkpoints, verification, and audit reports.

In plain words
What is it for?
Governing actions such as Git pushes and GitHub changes, recording authorization and results for higher-risk remote operations.
Why use it?
It makes agent work easier to authorize, check, reverse, and review, especially when external systems may be changed.

Cursor rule for Cursor

Written for Cursor: a Cursor rule (.mdc).

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/d-padmanabhan/agent-engineering-handbook/020-agent-audit
Clone the repo
git clone --depth 1 https://github.com/d-padmanabhan/agent-engineering-handbook

Made for: Cursor.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for 020-agent-audit

README.md
[![agentmods](https://agentmods.dev/badge/rules/d-padmanabhan/agent-engineering-handbook/020-agent-audit.svg)](https://agentmods.dev/rules/d-padmanabhan/agent-engineering-handbook/020-agent-audit)
Your own site
<a href="https://agentmods.dev/rules/d-padmanabhan/agent-engineering-handbook/020-agent-audit"><img src="https://agentmods.dev/badge/rules/d-padmanabhan/agent-engineering-handbook/020-agent-audit.svg" alt="Measured on agentmods" height="20"></a>
Per session 2,491 This file is loaded in full into every session.
When invoked 2,491 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 5.1 $0.02491 $0.02491
Opus 5 $0.01246 $0.01246
Sonnet 5 $0.00498 $0.00498
Haiku 4.5 $0.00249 $0.00249

Measured 6d ago against content hash 63695767b8ec, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

020-agent-audit 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 6d 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.

rules/020-agent-audit.mdc · 270 lines

How it starts

The opening of the file, as written. The whole thing — 270 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Workflow & Audit Requirements

These rules apply to AI agents operating in this workspace. They are designed to make work reversible, verifiable, and auditable.


1) Remote Mutations: Explicit Authorization; Read-Only Operations Allowed

Read-only operations do not require checkpoints or extra approval. Examples include file reads/searches, API GET requests, listing/status/describe commands, plans/diffs, CLI --help, and read-only MCP/tool calls.

Remote mutations require user authorization. A user's direct request to perform a specific mutation counts as authorization; do not ask for redundant approval unless the operation is destructive, irreversible, security-sensitive, or materially broader than the request.

For high-risk remote mutations, record:

  • the exact authorization (who/when/what),
  • the exact commands executed,
  • the results and exit codes, in the audit report.

This includes (non-exhaustive):

  • Git remote: git push, tag pushes, changing remotes, or any operation that writes to origin/upstream. Plain git fetch is read-only with respect to the remote.
  • GitHub remote: creating/merging PRs, pushing branches, forking repos, editing issues/PRs via write APIs.
  • Cloud / infra / data planes: terraform apply, kubectl apply, helm upgrade, aws cloudformation deploy, database migrations against non-local DBs, or any command that changes remote resources.

Commits are local-only and allowed only after explicit user authorization (see 130-git.mdc (${HANDBOOK_ROOT}/rules/130-git.mdc)).

When a push is authorized, fetch first and inspect branch divergence. Rebase or merge only when needed and permitted by repository policy; do not blindly run git pull --rebase against a dirty or shared branch. Use git push -u origin HEAD for a brand-new branch. See 130-git.mdc (${HANDBOOK_ROOT}/rules/130-git.mdc#branches-history-and-remote-writes). Never use --force or --force-with-lease unless the user explicitly authorizes that specific update.

Read the full file on GitHub · 270 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. 6d ago First seen · 270 lines · 2,491 tokens per session scan A 63695767b8ec

Subscribe to this mod's changes

020-agent-audit is a cursor rule published in the GitHub repository d-padmanabhan/agent-engineering-handbook (16 stars, last pushed 6d ago), licensed MIT. It adds 2,491 tokens to every session, about $0.0125 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.