audit

audit is a skill for Claude Code from sjh9714/red-handed. It costs 89 tokens per session (676 once invoked), scanned A, original, MIT.

A local checker that compares reported work with a Claude Code session transcript and the Git state. It checks whether claimed tests actually ran and whether test results or tests were altered.

In plain words
What is it for?
It is for auditing the latest or all project sessions, checking only the Git changes when no transcript exists, and producing readable or JSON findings.
Why use it?
It helps catch inaccurate claims about passing tests, changed expected values, or disabled tests before they are trusted.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: mentions Claude Code.

Part of the red-handed plugin — 2 skills, 2 hooks shipped together

Good fit It is for auditing the latest or all project sessions, checking only the Git changes when no transcript exists, and producing readable or JSON findings.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/sjh9714/red-handed/audit
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.

Any agent
npx skills add sjh9714/red-handed --skill audit
Clone the repo
git clone --depth 1 https://github.com/sjh9714/red-handed

Made for: Claude Code.

Or install red-handed, the plugin that ships this one along with the rest of its 2 skills, 2 hooks.

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 audit

README.md
[![agentmods](https://agentmods.dev/badge/skills/sjh9714/red-handed/audit/github.svg)](https://agentmods.dev/skills/sjh9714/red-handed/audit)
Your own site
<a href="https://agentmods.dev/skills/sjh9714/red-handed/audit"><img src="https://agentmods.dev/badge/skills/sjh9714/red-handed/audit/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for audit

Your own site · 80×15
<a href="https://agentmods.dev/skills/sjh9714/red-handed/audit"><img src="https://agentmods.dev/badge/skills/sjh9714/red-handed/audit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 89 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 676 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 $0.00089 $0.00676
Opus 5 $0.00044 $0.00338
Sonnet 5 $0.00018 $0.00135
Haiku 4.5 $0.00009 $0.00068

Measured 11d ago against content hash 573ff0fbb4eb, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

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 11d 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.

skills/audit/SKILL.md · 67 lines

How it starts

The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Audit this session

Run the audit and report what it found. Nothing here needs to be installed first; npx fetches the CLI on demand.

Running it

The most recent session for this directory:

npx --yes @jinhyuk9714/red-handed@latest audit

The subcommand is required. Running the tool bare reports how often "tests pass" had a real run behind it across the whole machine, which is a different question.

Useful variations:

  • --all — every session recorded for this project, not just the latest
  • --lang ko — report in Korean (it follows the system locale by default)
  • --git-only — when there is no transcript, audit the working tree diff instead
  • --json — when you need to read the findings programmatically rather than quote them

The command exits 1 when something reached the CAUGHT tier and 0 otherwise. A non-zero exit here is the tool working, not an error to retry.

Reporting what it found

Show the tool's own output. It is written to be read by a person: each finding already carries a plain-language description, the evidence with timestamps, and one thing to go check.

Then, briefly:

  • If nothing was found, say so plainly. That is the common case and it is not a disappointment. Do not pad it.
  • If something was found, state it without softening and without arguing. When the finding is about this session, it is about your own work — resist the pull to explain it away. The user asked what the record says; the record is what it says.
  • CAUGHT means the session shows it happening and the change is still in the working tree. SUSPICIOUS means the pattern is there but the motive is not established — an empty catch block is sometimes exactly right.
  • If a finding looks wrong to you, say why in one sentence and point the user at the quoted evidence so they can judge. Do not silently drop it.

What it cannot see

Worth saying out loud when it returns nothing, so a clean result is not mistaken for a guarantee:

  • Verification it cannot read counts as verification it did not see. A project's own test script, a browser check, anything without machine-readable output.
  • Claim sentences are matched in English, Korean, Japanese and Chinese. The report itself is written in English or Korean.
  • It says nothing about whether the code is correct. It only reports where the record does not support the claim.

Read the full file on GitHub · 67 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. 11d ago First seen · 67 lines · 89 tokens per session scan A 573ff0fbb4eb

Subscribe to this mod's changes

audit is a skill published in the GitHub repository sjh9714/red-handed (4 stars, last pushed 1mo ago), licensed MIT. It adds 89 tokens to every session and 676 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 2026-08-31.