market-intel

# market-intel A thin orchestration layer for commercial/market research. It does **only three things that nothing else does**: (1) triage a commercial topic to the right data domains, (2) detect which specialized MCP sources are actually connected and guide installing missing ones, (3) enforce research quality guardrails. The heavy lifting, fan-out search, fetching, adversarial verification, citation synthesis, is **delegated** to the existing `deep-research` harness or `research-lit` skill. Do not re-implement those. > **Design philosophy (governs all changes): root-cause design, not incremental patching**, change > the assumption underneath a problem, not the symptom on top. This thin-delegation shape, the > first-class browser route, and the monotonic anti-regression refresh all follow from it. Full > statement in the repo's `PHILOSOPHY.md`; every change must pass "does it fix the framing, or just > patch a symptom?" ## When to stop and delegate immediately Before doing anything, decide if this skill even applies: - **Single-fact lookup / quick query** → just use plain web search. Do not invoke this workflow. - **General web-only deep report** (no specialized commercial source needed) → delegate to `deep-research` and exit. - **Academic / scientific literature** → delegate to `research-lit` and exit. - **Needs a specialized commercial source** (X data, real e-commerce prices, market/finance feeds, on-chain data, SEO metrics, social sentiment, lead data) → continu...