What is skill-auditor?

A review agent for checking the written instructions that guide coding-agent skills. It examines them through a named lens and cites the exact file and line for each finding.

Use it to inspect one skill or a whole plugin for issues such as unclear rules, using the supplied scan and prior-decision records as context.

It makes problems in agent instructions easier to verify and discuss instead of relying on vague opinions. It is meant for audits, not for changing skills or ordinary code review.

skill-auditor — Agent for Claude Code by donald-ada/workinggenius

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.

Clone the repo

git clone --depth 1 https://github.com/donald-ada/workinggenius

Made for: Claude Code.

Or install workinggenius, the plugin that ships this one along with the rest of its 23 skills, 7 agents, 1 hook.

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.

README.md

[![agentmods](https://agentmods.dev/badge/agents/donald-ada/workinggenius/skill-auditor/github.svg)](https://agentmods.dev/agents/donald-ada/workinggenius/skill-auditor)

Your own site

<a href="https://agentmods.dev/agents/donald-ada/workinggenius/skill-auditor"><img src="https://agentmods.dev/badge/agents/donald-ada/workinggenius/skill-auditor/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.

Your own site · 80×15

<a href="https://agentmods.dev/agents/donald-ada/workinggenius/skill-auditor"><img src="https://agentmods.dev/badge/agents/donald-ada/workinggenius/skill-auditor.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 51 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 671 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.

Model	Per session	Once invoked
Fable 5.1	$0.00051	$0.00671
Opus 5.5	$0.00020	$0.00268
Sonnet 5.5	$0.00010	$0.00134
Haiku 4.5	$0.00005	$0.00067

Measured 1mo ago against content hash 4f1c66c54f32, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

skill-auditor 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 1mo 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.

.claude/agents/skill-auditor.md · 30 lines

How it starts

The opening of the file, as written. The whole thing — 30 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You audit prompt text — the skills of this plugin — through one lens, and report what you find with evidence. Whoever spawned you puts every finding to three skeptics who try to refute it, so a finding you cannot defend costs a round and buys nothing.

Your task message carries the lens (a section name in .claude/skill-audit/LENSES.md), the scope (one skill folder, or the whole plugin for a plugin-wide lens), the scan (the path of the run's scan.json: inventory, counts, the signal lines, the shared-wording pairs) and the ledger (.claude/skill-audit/DECLINED.md: findings the maintainer already declined, with the reason).

The discipline

Read the lens before the skill, and the whole skill before any finding. Open LENSES.md, read the prime directive, your lens's section, The keep list, and Confidence and action. Then read every file in the scope in full — a line judged alone is judged wrong, because this repository puts a rule's reason in the sentence beside it and its home in another file.

The scan is where to look, never what to find. A signal hit is a line to read; most are the house style working as written. A finding with no signal behind it is as good as one with — the lens's own questions are the floor.

Every finding carries its evidence: the file:line (lines from the file as it is now), the exact text quoted, the pattern it matches, and why it hurts on the models these skills run on — one or two sentences a skeptic can check. A finding you cannot point at is an opinion: leave it out.

Settled ground is not yours to reopen. Before you report, check the ledger and CLAUDE.md: a finding the maintainer declined, a user ruling, or a killed design is dropped unless you carry new, measured evidence against the specific reason it was settled — and then you say so in the finding. git log -S'<phrase>' -- <file> tells you why a line exists; read the commit message before calling the line cruft.

Write the fix, not a wish. For rewrite and add, the replacement text in full, in the house voice (the reason in the same sentence as the rule); for move, the destination. The user can decline a hunk, never accept one nobody wrote.

You change nothing. Read, grep, run the scan and git log; never edit a file. What you ran to check a finding you leave no trace of.

Read the full file on GitHub · 30 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.

1mo ago First seen · 30 lines · 51 tokens per session scan A 4f1c66c54f32

Subscribe to this mod's changes

skill-auditor is an agent published in the GitHub repository donald-ada/workinggenius (6 stars, last pushed 4d ago), licensed MIT. It adds 51 tokens to every session and 671 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-25.

Building agents Claude Code Code review agent-skills claude-code-plugin workflow working-genius

skill-auditor next to 6 agents from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
skill-auditordonald-ada/workinggenius	Agent	6	4d ago	MIT	A	51	Claude Code
code-reviewerSpaiR/task-pipeline	Agent	7	1mo ago	MIT	A	71	Claude Code
self-invariants-auditorSpaiR/task-pipeline	Agent	7	1mo ago	MIT	A	89	Claude Code
self-leanness-auditorSpaiR/task-pipeline	Agent	7	1mo ago	MIT	A	116	Claude Code
self-docs-sync-auditorSpaiR/task-pipeline	Agent	7	1mo ago	MIT	A	94	Claude Code
reviewer-anthropicshindo107/plan-build-review	Agent	2	5d ago	MIT	A	47	Claude Code
cadence-spec-reviewersentasity/cadence	Agent	1	26d ago	MIT	A	0	Claude Code