{"name":"Outlet Profiler","agent_id":"outlet-profiler","description":"Profiles every outlet by opportunity — how well it performs versus structurally-similar peers on size-neutral intensity levers — and assigns an opportunity tier (T1–T4) plus a grade-as-vector per business play. A CPG-OS Case-C depth agent: it OWNS the outlet opportunity-tier ground truth other products defer to, validates supervisor hypotheses about that truth data-first, and never orchestrates or sets cross-product priority. Every response is tagged deterministic (a rule over the data) or reasoning (an interpretation call).","url":"http://localhost:8100/a2a","version":"0.1.0","capabilities":{"streaming":false,"pushNotifications":false},"defaultInputModes":["application/json"],"defaultOutputModes":["application/json"],"securitySchemes":{"bearer":{"type":"http","scheme":"bearer"}},"security":[{"bearer":[]}],"produces_contract_types":["observation","diagnosis","opportunity"],"accepts_contract_types":["diagnosis","opportunity"],"reasoning_modes":["deterministic","reasoning"],"skills":[{"id":"grade_outlets","name":"Grade outlets by opportunity","description":"Opportunity tier + grade-vector + ₹-sized headroom per outlet. A plain-English play is parsed into a grading lens by an optional Claude LLM lens (reasoning) with a deterministic keyword fallback; built-in plays include premium_launch, volume_scheme, frequency (order-cadence), distribution, retention, reactivation, balanced."},{"id":"validate_opportunity_hypothesis","name":"Validate a hypothesis","description":"Confirm/refute/inconclusive verdict on a supervisor assertion, data-first."},{"id":"analyze_outcome","name":"Measure an outcome","description":"Post-action impact into the learning loop (stub)."}]}