research-prospect

name: research-prospect description: "Run ad-hoc web research on a prospect (company URL or person) — Firecrawl scrapes the marketing site, key pages (pricing, careers, blog), and surfaces a brief. Read-only. Use when the user says 'research this prospect', 'tell me about [company]', 'do a deep dive on [domain]', 'pull a brief on this lead', or 'crawl the marketing site for [name]'. Side-effecting on the Firecrawl side (API call), no local DB writes." version: 1.0.0

Research Prospect

I'll wrap research. Take a URL, run a Firecrawl scrape against the canonical pages, and produce a brief.

When This Skill Applies

• "research this prospect"
• "tell me about [company]"
• "do a deep dive on [domain]"
• "pull a brief on this lead"
• "crawl the marketing site for [name]"

NOT this skill (use enrich-with-signals instead):

• "pull buying signals" — that's structured PredictLeads data, not narrative research.

NOT this skill (use personalize-message with --enrich instead):

• "personalize a message for them" — that auto-pulls research as part of the personalization.

Workflow

Step 0 — Ask for the URL

"What's the company URL or LinkedIn profile?"

Step 1 — Validate URL

Step 2 — Shell out

cd ~/Desktop/gtm-os && set -a && source .env.local && set +a && \
  npx tsx src/cli/index.ts research --url <url>

Step 3 — Parse output

CLI emits a brief markdown summary + key facts table.

Step 4 — Render the brief verbatim

Don't summarize the summary — pass through what the CLI returns.

Step 5 — Offer follow-ups

"Want me to (a) personalize a message for someone at this company via personalize-message, (b) check buying signals via enrich-with-signals?"

Notes

• Requires FIRECRAWL_API_KEY in ~/.gtm-os/.env.
• Firecrawl call cost varies by depth (default: 1 credit per ~5 pages).
• Read-only on the local DB — research output is rendered in chat, not persisted.