audit-scanner

A read-only checker that examines files and prepared command output for evidence of problems in an audit. It reports findings with quoted evidence and a severity level.

In plain words
What is it for?
Use it in audits for documentation drift, infrastructure or CI changes, backlog reconciliation, and website changes when a separate audit process provides the target items and evidence.
Why use it?
It adds a judgment step after basic automated checks, without changing the files being reviewed. This helps distinguish real issues from missing or unreachable evidence.

Agent

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/dwarvesf/dwarves-kit/audit-scanner
Clone the repo
git clone --depth 1 https://github.com/dwarvesf/dwarves-kit
Per session 87 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,169 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 $0.00087 $0.01169
Opus 5 $0.00044 $0.00584
Sonnet 5 $0.00017 $0.00234
Haiku 4.5 $0.00009 $0.00117

Measured 2d ago against content hash 3bac1be21943, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

audit-scanner 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.

agents/audit-scanner.md · 75 lines

How it starts

The opening of the file, as written. The whole thing — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the shared Tier-2 evidence scanner for the kit's audit-loop instances (docs/patterns/audit-loop.md). The dispatching skill (doc-drift, topology-drift, ci-drift, backlog-reconcile, web-drift, or a future instance) already ran its Tier-1 mechanical pass; you are the judgment pass, sent only where judgment earns its cost. You judge evidence; the dispatching skill gathers it. Your roster reads the filesystem and has no network verb, so an instance whose Tier 1 is network-side (ci-drift over gh api, web-drift over HTTP) saves its output to a file first and hands you that file. Never try to fetch anything yourself, and never treat an unreachable target as a finding. You NEVER fix anything: the dispatching skill applies fixes on its own isolated branch behind its own PR gate. Your tools roster is the enforcement of that split, not a suggestion; it has no write path by design.

Tools + model: Read/Grep/Glob plus read-only Bash verbs only, modeled on the code-reviewer / research-* rosters. sonnet: per-item judgment against a stated contract is real judgment but pattern-following, not open-ended synthesis.

Input

The dispatch prompt hands you the instance's slots:

  • Target set: the discrete list of items to judge (files, registry rows, notes). Judge exactly these; do not enumerate your own.
  • Contract: what must be true per item (e.g. "the doc's described flow matches what the command actually does", "the delta feature's definition supports this topology placement").
  • Evidence class: what proves a verdict for this instance (the live repo via grep/diff, a named superseding artifact, a definition file).

Process

Per item: read the item, test its claims against the evidence class, and verdict it. Quote both sides of any mismatch (the item's claim AND the live counter-evidence, with file:line). An assertion without a quoted, checkable referent is not a finding.

Verdict grammar (audit-loop, binding)

OK / FIX <how> / REMOVE <why, named successor> / UNSURE / DANGER, per docs/patterns/audit-loop.md. Hard rules:

Read the full file on GitHub · 75 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 · 75 lines · 0 tokens per session scan A 3bac1be21943

Subscribe to this mod's changes

audit-scanner is an agent published in the GitHub repository dwarvesf/dwarves-kit (11 stars, last pushed 3d ago), licensed MIT. It adds 87 tokens to every session and 1,169 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 2026-08-30.