ai-redteam
llms.txt2 starsChanged 9 days agoMITView on GitHub
Adversarial review for AI-produced work: a redteam skill (decisions, code, claims, ideas, regulated text), a pr-review skill and an assess skill (is a link useful here?), with paste-in prompts, reference docs and a scored eval set. Agents: start at AGENTS.md. People: start at README.md.
The file
Files
# ai-redteam > Adversarial review for AI-produced work: a redteam skill (decisions, code, claims, ideas, regulated text), a pr-review skill and an assess skill (is a link useful here?), with paste-in prompts, reference docs and a scored eval set. Agents: start at AGENTS.md. People: start at README.md. ## Skills - [skills/redteam/SKILL.md](skills/redteam/SKILL.md): adversarial review; Tracks A decisions, B code, C claims, D ideas, R regulated text; inputs ledger, sensitivity gate, confirm-or-refute - [skills/pr-review/SKILL.md](skills/pr-review/SKILL.md): bounded review of one pull request with written adjudication - [skills/assess/SKILL.md](skills/assess/SKILL.md): whether a link or item is useful here, its claims checked, and one next action ## Prompts (no install) - [prompts/adversarial-review.md](prompts/adversarial-review.md): redteam as a paste-in prompt for any model - [prompts/pr-review.md](prompts/pr-review.md): pr-review as a paste-in prompt - [prompts/assess.md](prompts/assess.md): assess as a paste-in prompt for chat assistants - [prompts/cross-disciplinary-review.md](prompts/cross-disciplinary-review.md): review a claim across disciplines, prior work first (prompt only, no skill) - [templates/assess-context.md](templates/assess-context.md): the blank context file assess reads (keep a filled copy private) ## Docs - [docs/README.md](docs/README.md): documentation index - [docs/eval-requests/README.md](docs/eval-requests/README.md): independent evaluation requests - [AGENTS.md](AGENTS.md): rules and commands for AI agents working in this repo - [docs/capabilities.md](docs/capabilities.md): everything the repo does, grouped, each with the file that owns it - [docs/using-with-ai-agents.md](docs/using-with-ai-agents.md): how each CLI reads a repo, install and invoke, proving the load, pitfalls - [docs/workflow.md](docs/workflow.md): where these reviews sit in a build-and-review loop - [docs/SPEC.md](docs/SPEC.md): redteam design and failure list, version by version (the file owns its versions and item numbers) - [docs/attack-catalog.md](docs/attack-catalog.md): ways AI-built systems fail under attack, as reviewer questions - [docs/framework-mapping.md](docs/framework-mapping.md): the catalog and skills against the OWASP LLM and Agentic Top 10 lists and MITRE ATLAS - [docs/security-audit-mapping.md](docs/security-audit-mapping.md): Cloudflare security-audit findings.json mapped into the findings schema - [docs/why-reviews-fail.md](docs/why-reviews-fail.md): how reviews of AI work go wrong - [docs/privacy-checklist.md](docs/privacy-checklist.md): questions for work that touches personal data ## Evals - [evals/README.md](evals/README.md): baseline suite of 92 cases plus dedicated suites; scorer, runner and scoring limits - [evals/results/](evals/results/): published runs with reports, prompts and SHA256SUMS - [docs/variance-2026-10-16.md](docs/variance-2026-10-16.md): dated descriptive variance, not new model runs - [docs/pr-review-action.md](docs/pr-review-action.md): opt-in Action, deterministic OSV lookup and bounded model round
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.