market-researcher

A Claude Code skill that decides whether to build a SaaS idea.

Runs hybrid research (parallel Haiku subagents + human-in-the-loop), produces a tier-adjusted scorecard against 12 criteria, and outputs a folder of cited markdown reports plus a 15-slide Reveal.js HTML deck with a clear GO / PIVOT / KILL verdict.

Install

Clone into your Claude skills directory:

git clone <this-repo-url> ~/.agents/skills/market-researcher

Then invoke from Claude Code with the skill picker (or by asking "should I build a SaaS for X?").

Files

• SKILL.md — entrypoint loaded by Claude
• references/ — sources registry, subagent prompts, scorecard rubrics, output templates, and the canonical Reveal.js slide template
• examples/ — worked example showing a complete output folder
• DESIGN.md / PLAN.md — design notes and implementation plan

Output

Each run produces research/YYYY-MM-DD-<idea-slug>/ containing:

• context.md — the intake brief so a future chat can resume cold
• 00-verdict.md — TL;DR, scorecard, verdict, top 3 risks, validation checklist
• 01-market-sizing.md … 08-sources.md — the cited research files
• slides.html — a self-contained Reveal.js deck (open ?print-pdf in Chrome to export PDF)