Agent-access policy conformance tester
Publishing a policy is only half the job — an agent or tool has to be able to find it and read it. This tester acts as an independent consumer: it fetches /.well-known/agent-policy.json from the site you name, unauthenticated, parses it with a fresh consumer-side reader built from the published AGENT-POLICY v0.1 spec, and reports — check by check — what it discovered, what it understood, and what it could not interpret.
Publishing ≠ enforcement
AgentGate publishes policy and this tester checks discoverability + interpretation. Nothing here enforces a policy: actually stopping an agent that ignores one happens at a gateway (for example Agent Auth Gateway), and this page does not claim it. A conformant result means a consumer could find and read the policy — never that any agent will obey it.
What gets checked
- Discovery — the well-known URL answers HTTP 200, without authentication, as JSON.
- Parsing — the body is valid JSON with the document shape the spec defines.
- Interpretation — every field is read by the consumer parser and shown with the meaning the consumer derived: the baseline decision, each agent rule's decision, actions, terms and rate limit.
- Uninterpretable fields — anything outside the v0.1 spec is flagged by name, not silently dropped.
Free and signed-out, like the generator and validator. Fetches are SSRF-guarded, timed out at 8 seconds and capped at 256 KB, as the spec requires of any remote fetch.