browser-researchlisted

# Browser Research This Skill fills the exact capability gaps documented in `references/provider-capability.md` — no automated provider in this project has a live review-text API, discovers by free-text concept reliably, or looks outside Amazon at all. It does not replace the deterministic engine; it feeds evidence into it, at the engine's own defined tier, through the engine's own commands. ## 0. Precondition check — do this first, every time Browser research is expensive (your time, and the human's time reviewing results). It is reserved for candidates that have already survived cheap discovery, cheap filtering, economics check, and competitor autopsy. Run `hunt show <opportunity_id>` and confirm `state` is one of: `COMPETITOR_ANALYSIS`, `VOC_ANALYSIS`, `DIFFERENTIATION_ANALYSIS`, `RED_TEAM`, `HUMAN_REVIEW`, `QUALIFIED`, `PENDING_HUMAN_REVIEW`. This is the exact same `ELIGIBLE_STATES` allow-list `hunt browser-brief` itself enforces (I3 in the rule inventory) — it will refuse (`ValueError`, not a silent no-op) below `COMPETITOR_ANALYSIS`. **Do not attempt this Skill on a raw `DISCOVERED`/`PRE_FILTERED`/`VALIDATING`/`ECONOMICS_CHECK` candidate** — the engine's own refusal exists precisely because running this against the full discovery pool (often 100+ candidates) would violate cost discipline (`core-rules.md`, time/cost section). ## 1. Step one — generate the brief ``` hunt browser-brief <opportunity_id> ``` This prints (and by default saves) an **EXTERNAL BROWSER RES