Market Research Skill

market-research — a Claude skill for finding your customers' own words

Mine what people actually say online about a market — their pain points, the phrases they use, the tools they've abandoned — and turn it into a positioning brief.

**Thesis:** the best-positioned product wins, not the "best" product. There are already customers on Reddit / X / TikTok / YouTube talking about the problem your product solves, in their own words. Find that language, cluster it, and feed it back to them on your landing page, your onboarding, your cold emails — and the marketing lands with no warm-up. Some people call this *content–market fit*, and it comes before product–market fit.

This skill automates the listening half: it sweeps the platforms, ranks by real engagement, **transcribes the videos** (where people describe problems out loud and the caption doesn't), dedupes, and hands you one corpus. You (Claude) do the synthesis — the ranked pain-point list, positioning angles, competitor/gap map, and ICP language bank.

What it is

A single Claude Code / Claude.ai **skill** — not a multi-agent system. Claude follows `SKILL.md` and calls two helper scripts as tools:

File	Role
SKILL.md	The 6-step playbook Claude follows (scope → pre-research → collect → transcribe → synthesize → deliverables).
scripts/collect.py	Wraps the last30days engine: pain-mining query plan + disambiguated competitor sweep → merged corpus.md / corpus.json.
scripts/transcribe.py	Downloads video audio (yt-dlp) and transcribes it locally (Whisper), folds transcript: lines back into the corpus.
references/	deliverables.md (output spec), pain-taxonomy.md (how to cluster), query-plan.md (plan schema).

The scripts are deterministic data plumbing — no LLM calls inside them. All reasoning is the one Claude loop reading `SKILL.md` and the corpus.

Install

1. The skill

git clone https://github.com/teeyoone/market-research-skill.git ~/.