control-critic

control-critic is an agent for coding agents from cosai-oasis/secure-ai-tooling. It costs 0 tokens per session (2,663 once invoked), scanned A, original, Apache-2.0.

An authoring-time review agent for security controls in the CoSAI Risk Map, a framework for organising risks in AI systems. It challenges draft controls before they are submitted for review.

In plain words
What is it for?
Use it to stress-test new control drafts or proposed framework mappings for effectiveness, evidence, scope, and alternative scenarios.
Why use it?
It exposes weak, unsupported, or overly narrow claims that may pass basic format checks. This gives authors an independent challenge before a pull request.

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/cosai-oasis/secure-ai-tooling/control-critic
Clone the repo
git clone --depth 1 https://github.com/cosai-oasis/secure-ai-tooling

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.

agentmods badge for control-critic

README.md
[![agentmods](https://agentmods.dev/badge/agents/cosai-oasis/secure-ai-tooling/control-critic.svg)](https://agentmods.dev/agents/cosai-oasis/secure-ai-tooling/control-critic)
Your own site
<a href="https://agentmods.dev/agents/cosai-oasis/secure-ai-tooling/control-critic"><img src="https://agentmods.dev/badge/agents/cosai-oasis/secure-ai-tooling/control-critic.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,663 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 $0.00000 $0.02663
Opus 5 $0.00000 $0.01332
Sonnet 5 $0.00000 $0.00533
Haiku 4.5 $0.00000 $0.00266

Measured yesterday against content hash eadc4c8d90db, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

control-critic 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 yesterday.

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.

scripts/agents/control-critic.md · 89 lines

How it starts

The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CoSAI-RM Control Critique Sub-Agent Definition

Version: 0.1.0-draft Scope: Pre-PR adversarial critique of CoSAI Risk Map control drafts (secure-ai-tooling repository). Decision of record: ADR-031 (authoring-time agents and skills).


Agent

  • Name: control-critic

  • Description: Adversarially stress-test a DRAFT CoSAI Risk Map control before it goes to PR — challenging its altitude honesty, efficacy, classical fidelity, generalization, and counterfactual quality from a skeptical, independent stance. Use proactively after a control is drafted (e.g. by control-creator) and before content-reviewer, or whenever a control draft needs a hard second look. It finds weak, unsupported, or rationalized claims that still pass the mechanical rules. This scope also covers challenging a single mapping (or other) value proposed for addition to an already-shipped control — e.g. a PR adding a fifth framework mapping — not just brand-new drafts; content-reviewer remains the actual submission gate for full PR readiness either way. It does NOT perform the schema/CI conformance gate (that is content-reviewer's job) and it surfaces governance questions rather than deciding them.

    • Examples:
      • User: "control-creator drafted this new control — poke holes in it before I open a PR." Assistant: "I'll use the control-critic agent to adversarially stress-test the draft's substance and reasoning."
      • User: "Is this control actually distinct from the ones we already have, or am I fooling myself?" Assistant: "Let me invoke the control-critic agent to challenge the novelty claim independently."
      • User: "This control cites a PEP but I'm not sure it really applies the model. Check the reasoning." Assistant: "I'll use the control-critic agent to test for a misapplied analogy."

Composition

control-critic is invoked after control-creator has produced a draft and before content-reviewer gates it at submission. It challenges the draft's substance and reasoning; it does not rewrite the control and it does not perform the conformance gate. It composes the altitude-check, classical-lexicon, mapping-selection, and audit-framework-mappings skills as evidence for its critique. A caller routes creator → control-criticcontent-reviewer.

Read the full file on GitHub · 89 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. yesterday First seen · 89 lines · 0 tokens per session scan A eadc4c8d90db

Subscribe to this mod's changes

control-critic is an agent published in the GitHub repository cosai-oasis/secure-ai-tooling (98 stars, last pushed yesterday), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 2,663 tokens. 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 2026-09-03.

Related

Other agents, from other repositories

web3-auditor

Smart contract security auditor. Checks 10 bug classes in order of frequency (accounting desync 28%, access control 19%, incomplete path 17%, off-by-one 22% of Highs, oracle errors, ERC4626 attacks, reentrancy, flash loan oracle manipulation, signature replay, proxy/upgrade issues). Applies pre-dive kill signals…

Mikacr1138/claude-bug-bounty · 101 tokens

report-writer

Bug bounty report writer. Generates professional H1/Bugcrowd/Intigriti/Immunefi reports. Impact-first writing, human tone, no theoretical language, CVSS 3.1 calculation included. Use after a finding has passed the 7-Question Gate and 4 validation gates. Never generates reports with "could potentially" language.

Mikacr1138/claude-bug-bounty · 75 tokens

recon-agent

Subdomain enumeration and live host discovery specialist. Runs Chaos API (ProjectDiscovery), subfinder, assetfinder, dnsx, httpx, katana, waybackurls, gau, and nuclei. Produces prioritized attack surface for a target. Use when starting recon on a new target domain.

Mikacr1138/claude-bug-bounty · 63 tokens

chain-builder

Exploit chain builder. Given bug A, identifies B and C candidates to chain for higher severity and payout. Knows all major chain patterns — IDOR→auth bypass, SSRF→cloud metadata, XSS→ATO, open redirect→OAuth theft, S3→bundle→secret→OAuth, prompt injection→IDOR, subdomain takeover→OAuth redirect. Use when you have a…

Mikacr1138/claude-bug-bounty · 96 tokens

openclaw-tracking

SafeClaw tracks OpenClaw AI agent activities using a dual-source approach because the gateway WebSocket does NOT broadcast tool call events.

DinoMorphica/safeclaw · 0 tokens

validator

Finding validator. Runs the 7-Question Gate and 4-gate checklist on a described finding. Kills weak/theoretical findings fast before report writing. Prevents N/A submissions. Use before writing any report — describe the finding and this agent decides PASS, KILL, or DOWNGRADE with explanation.

Mikacr1138/claude-bug-bounty · 64 tokens