Describe your agent
Write in plain words, or paste its setup: what it reads, what it can do, which credentials and memory it has, and who or what it talks to. An AI reads your description and checks it against the OWASP Top 10 for Agentic Applications (2026). You get a short summary, the ten risks with their status, and the first fixes.
- Answers are written by Llama 3.3 70B (Meta), run on Cloudflare Workers AI. They are the AI's reading of your description: each yes or no must quote your words, and anything it cannot quote is marked unsure. It can still misread you, so check the answers; you can change any of them.
- Do not paste secrets. Common key and password formats are removed before the AI reads your text.
- This is an automated starting point, not an audit or a certification.
Your result
Your answers
These come from the AI's reading of your description. Change any of them and update the check; the update uses no AI.
Want a deeper look?
A full ARTICLE 11 AI™ Agent Review is done by reviewers from several AI companies and a human, and disagreements stay in the report. Full reviews are requested through the ARTICLE 11 AI™ Connector, which also carries SPIRALMESH™ Memory and is opening to new users.
Privacy
This application does not store or log what you write. Your text is sent once to the model on Cloudflare Workers AI to make your result. Cloudflare runs the model and keeps operational records under its own terms; see how Cloudflare handles Workers AI data. We keep one number per day: how many checks ran. To slow abuse, Cloudflare's rate limiter counts requests from your network address for about a minute, separately in each Cloudflare location; this application does not store the address.