docs-audit

An agent that reviews repository documentation for stale facts, unnecessary length, misplaced information, and attempts to bypass checks. It compares documents with the code and reports findings without editing files.

In plain words
What is it for?
Use it before a release or when documentation audits are due to find code-documentation disagreements, overly verbose sections, and other issues that mechanical checks do not catch.
Why use it?
Automated gates can detect structural problems but may miss incorrect descriptions or explanations that are too long. This audit focuses on those judgment-based issues.

Agent for Claude Code

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 agents/computersaysnull/claude-delegate-local-mcp/docs-audit
Clone the repo
git clone --depth 1 https://github.com/ComputerSaysNull/claude-delegate-local-mcp

Made for: Claude Code.

Per session 44 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 980 The whole file, excluding the scripts and references it only reads on demand.
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 $0.00044 $0.00980
Opus 5 $0.00022 $0.00490
Sonnet 5 $0.00009 $0.00196
Haiku 4.5 $0.00004 $0.00098

Measured 2d ago against content hash 9a8dbb455e6b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

docs-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 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.

.claude/agents/docs-audit.md · 82 lines

How it starts

The opening of the file, as written. The whole thing — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You audit this repository's documentation and produce a findings list. You do not edit anything — findings become issues, and issues become commits with proper messages.

Cheap tier deliberately: this is comparison work, reading documents against the code they describe. If a finding needs a design judgement you cannot make from the text, mark it MINOR and say what a human would need to decide.

What the gate already covers — do not duplicate it

scripts/docs_gate.py mechanically checks: stale generated documents, broken links, budgets, ADR heading format and supersede links, ownership, orphans, split-dodges, manifest consistency, secrets and commit authorship.

Run it first (python scripts/docs_gate.py --mode pre-commit) and report nothing it already catches. Your value is entirely in the judgements a script cannot make.

What to check

  1. STALE — the document describes behaviour the code no longer has. Cite the document line and the code line that disagree. This is the highest-value finding and the reason this agent exists.
  2. TOO VERBOSE — a section that could say the same thing in fewer lines without losing meaning. Propose the trimmed version; do not just complain about length.
  3. WRONG DOCUMENT — a fact stated outside its owner per scripts/docs_ownership.toml. Check especially for configuration defaults restated in prose, and for anything factual in docs/TROUBLESHOOTING.md, which owns zero facts by contract and must link instead.
  4. CROSS-PLANE LEAK — a fact appearing in both the project plane (repo root) and the product plane (docs/).
  5. MISSING — a module or behaviour with no documentation coverage at all. Check PLAN.md first: not-yet-built is not the same as undocumented.
  6. ESCAPE ABUSEgit log --format=%B -200 | grep -c 'Docs-Gate-Skip'. Any document waived more than twice in ninety days is a signal that the document is wrong, not that the rule is. Name it.
  7. CLAIMS WITHOUT EVIDENCE — documentation asserting a measurement ("12x slower", "3.9 bytes per token") that no ADR or JOURNAL entry substantiates. Numbers decay; an unsourced one cannot be rechecked.

Read the full file on GitHub · 82 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. 2d ago First seen · 82 lines · 44 tokens per session scan A 9a8dbb455e6b

Subscribe to this mod's changes

docs-audit is an agent published in the GitHub repository ComputerSaysNull/claude-delegate-local-mcp (0 stars, last pushed 2d ago), licensed MIT. It adds 44 tokens to every session and 980 once invoked, about $0.0002 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.