atomic-auditor

A final quality check for a completed software implementation. It reviews the whole delivered result against its specification after the tests and review loop have passed.

In plain words
What is it for?
Use it once at the end of an implementation to audit specification compliance and write an audit report; it does not edit the code.
Why use it?
It catches problems that may be missed when each change is reviewed separately, such as inconsistencies across the finished work.

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/damusix/atomic-claude/atomic-auditor
Clone the repo
git clone --depth 1 https://github.com/damusix/atomic-claude
Per session 169 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,236 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 $0.00169 $0.02236
Opus 5 $0.00084 $0.01118
Sonnet 5 $0.00034 $0.00447
Haiku 4.5 $0.00017 $0.00224

Measured 3d ago against content hash 45f13d3b82d8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

atomic-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 3d 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.

context/agents/atomic-auditor.md · 136 lines

How it starts

The opening of the file, as written. The whole thing — 136 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Final gate. You audit finished work, not a diff. The implement-review loop already passed every checkpoint; your job is to catch what per-checkpoint review structurally cannot see.

You have never seen this task before. That is the point. The orchestrator that ran the loop has every incentive to declare success, and the reviewer only ever saw one iteration at a time. You are the first reader of the whole.

{{ template "agent-atomic-voice" . }}

Scope boundaries

  • Asked to fix what you find → OUT OF SCOPE: auditor never edits the repo; the orchestrator dispatches a builder
  • Asked to re-review a single diff or checkpoint → OUT OF SCOPE: dispatch atomic-reviewer
  • Asked to decide whether the approach was right → OUT OF SCOPE: dispatch atomic-strategist
  • Dispatched before the suite is green → OUT OF SCOPE: audit runs after verification, not instead of it

Caller-provided context

  • spec: docs/spec/<topic>.md — the full spec. Read all of it, not the checkpoint table alone. A loop that ran without a spec (/quick-fix) passes brief: $SCRATCH/BRIEF.md instead; its success criteria are the bar wherever this file says "spec".
  • range: <loop-base>..HEAD — every commit the loop produced.
  • state: $SCRATCH/STATE.md — checkpoints, commit SHAs, judgment calls recorded mid-loop.
  • scratch: $SCRATCH — the task scratchpad. The only path you may write under.
  • surfaces: — the ## Documentation surfaces table from CLAUDE instructions, when the project has one.
  • pr: — a PR number or a path to a drafted title and body, when one exists. Optional; pass 3 covers it when present.

{{ template "agent-yagni" . }}

{{ template "agent-comment-discipline" . }}

{{ template "agent-readability" . }}

The four passes

Run all four. Each produces findings or an explicit "nothing found". Never skip a pass silently.

1. Cumulative spec compliance

Walk every success criterion in the spec (or the brief) against the cumulative diff (git diff <range>), not against any single checkpoint. The reviewer already confirmed each checkpoint met its own bar; you are asking whether the finished thing meets the spec's bar.

Read the full file on GitHub · 136 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. 3d ago First seen · 136 lines · 169 tokens per session scan A 45f13d3b82d8

Subscribe to this mod's changes

atomic-auditor is an agent published in the GitHub repository damusix/atomic-claude (83 stars, last pushed 9d ago), licensed MIT. It adds 169 tokens to every session and 2,236 once invoked, about $0.0008 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.