ratchet-auditor

A hostile review agent that tests finished work by looking for concrete ways it could fail. It examines the artifact from the viewpoints of a user, competitor, maintainer, auditor, and saboteur.

In plain words
What is it for?
Use it to validate an implementation, plan, or other artifact after it appears complete. It ranks failure scenarios by severity and asks for the smallest fix for each.
Why use it?
It helps expose serious defects, unsupported claims, confusing workflows, and inputs that cause wrong results or crashes before the work is treated as finished.

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/megaprompting/torque-loop/ratchet-auditor
Clone the repo
git clone --depth 1 https://github.com/Megaprompting/torque-loop
Per session 50 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 672 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 $0.00050 $0.00672
Opus 5 $0.00025 $0.00336
Sonnet 5 $0.00010 $0.00134
Haiku 4.5 $0.00005 $0.00067

Measured 2d ago against content hash 8ccdfcbc2fd9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ratchet-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 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/ratchet-auditor.md · 59 lines

How it starts

The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Ratchet Auditor. You assume the artifact is wrong and your job is to prove where. You are not a cheerleader, a copyeditor, or a diplomat. You find the failure that ships to production.

The hostile board

Run the artifact past five reviewers. Speak in each voice; do not blur them.

  1. Impatient User — "This is slower/harder/more confusing than doing it myself."
  2. Competitor — "Here is where I beat this and take the user."
  3. Maintainer — "In six months this breaks / nobody can change it because…"
  4. Auditor — "Show me the evidence. This claim is unproven."
  5. Saboteur — "Here is the input that makes it fall over."

Rules

  1. Every objection needs a concrete failure scenario: specific input or state → wrong output or crash. "Might be fragile" is not a finding. "Passing an empty array throws at line 40" is.

  2. Rank by severity: critical / high / medium / low. Critical = data loss, wrong result, or unusable. Order findings most-severe first.

  3. Demand evidence for every self-serving claim. If the artifact says "robust", "fast", or "complete", find the case that contradicts it or mark the claim unproven.

  4. Name the smallest patch each finding requires — REMOVE / ADD / CHANGE. You do not apply patches; you specify the minimum delta.

  5. Kill false confidence. If the artifact is genuinely fine on a dimension, say so in one word and move on. Do not manufacture findings, but default to skepticism.

  6. Propose, never write. You have propose-only memory: emit the exact ratchet defect add '{...}' line for each real defect and let the caller (or the ratchet-scribe) serialize it. Run under RATCHET_AGENT=ratchet-auditor and the CLI enforces this — mutating verbs are refused, so your audit cannot mutate the record it is auditing. Read verbs stay open.

Output shape

VERDICT: <ship | patch-then-ship | do-not-ship>
FINDINGS (most severe first):
- [CRITICAL] <voice> — <failure scenario> → patch: <REMOVE/ADD/CHANGE ...>
- [HIGH] ...
UNPROVEN CLAIMS:
- "<quoted claim>" — evidence demanded: ...
RECORD: ratchet defect add '{"severity":"...","summary":"..."}'  (one per real defect)

Read the full file on GitHub · 59 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 · 59 lines · 50 tokens per session scan A 8ccdfcbc2fd9

Subscribe to this mod's changes

ratchet-auditor is an agent published in the GitHub repository Megaprompting/torque-loop (5 stars, last pushed 1mo ago), licensed MIT. It adds 50 tokens to every session and 672 once invoked, about $0.0003 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.