stage-1-competitivelisted

Stage 1: competitive & alternatives scan. Load `method-rules`; read `state.json`. Runs immediately after 0.1 — its output shapes stage 2's questions. Templates: the `stage-1-competitive-templates` skill. ## Execution 1. **Spawn agents in parallel** (Agent tool, one message): - `competitor-scanner` — with the problem statement, segment, and JTBD keywords from `problem-hypothesis.md`. Produces the 5-tier map + profiles + market verdict. - `community-review-miner` (Task B) — once the scanner returns significant tier-1/2 competitors, mine their 1–3★ reviews into unmet-need clusters with verbatim quotes. (If competitors are already obvious from stage 0, launch both at once.) Pass both agents `state.research_budget.per_task` if set (absent → plugin defaults). Both agents now consult `source-registry.md` before fetching and report a `Research budget` block (v1.4.0) — see below. 2. Rung note: agents use scraping MCP when `capabilities.scraping = available`; otherwise native search. Directory sites (G2/Capterra) often block plain fetches — collect the agents' "blocked sources" lists and, if material, offer the **handoff rung**: give the user exact URLs to open and paste raw content back for clustering. 3. Write `competitive-map.md` (frontmatter `status: draft` — it stays a DRAFT through gate C by design, until stage 2 calibrates it against real customer words and promotes it) and `review-mining.md` (**promote to `status: ready` once clustering is complete** — gate C require