GET /.well-known/chaos

Parametric chaos for well-known discovery schemas. Pick a schema (openid-configuration, oauth-authorization-server, webfinger, jwks, host-meta, agent-card) and a mode (semantic, missing-fields, wrong-types). The response is served with the correct Content-Type for the chosen schema, so clients that key on Content-Type behave as they would against a real server. Set ?ai=true to have an edge LLM generate a fresh source on each call — the mode transformation is then applied on top.

schema Which schema shape to base the response on. One of: openid-configuration, oauth-authorization-server, webfinger, jwks, host-meta, agent-card. Default: openid-configuration.

mode How to misbehave. semantic = base drift (URLs unreachable, claims fake). missing-fields = strip required fields. wrong-types = change one field's type. Default: semantic.

ai If true, source data is AI-generated fresh per call (different agent or issuer each time), then the mode transformation is applied. Same fallback behaviour and X-Chaos-Ai-* headers as /semantic-drift?ai=true. Default: false.

build a request:

schema mode ai-generated

expect: A discovery document of the chosen schema, transformed according to mode. Content-Type matches the schema's real-world type so Content-Type-keyed clients behave normally.

bash