Tier 2 · Hunt — build a repeatable method.
Map the real surface; find what the docs don't mention.
/method-recon/Build a role×object matrix; find the cells that break.
/method-authz/Catalog every token; find the one that's weak.
/method-tokens/Enumerate every input; find the one that's unsafe.
/method-injection/Diff fields in vs out; find over-exposure & mass-assign.
/method-dataflow/Chain minor findings into account takeover.
/method-chaining/## Methodology section, follow the numbered steps, and keep notes (an endpoint list, an input catalog,
a role×object matrix). The decoys are there to teach you to rule things out with evidence. When a real finding
surfaces, a CAPIE{} flag confirms it.