Back to work

Signal Scout

Evidence-backed go-to-market research agent · AI Agents

A Claude Code / OpenCode skill that turns a startup URL into an evidence-backed shortlist of first customers, market segments, and companies to pitch, using public signals only.

View source on GitHub

project page

how it was built

Problem

LLM-generated market research reads well but is routinely wrong: a model will confidently name a "prospect" or cite a "signal" that isn't actually there when you go look. For go-to-market research specifically, a fabricated lead costs real outreach time before anyone notices.

Decisions

  • Built verify_sources.py as a hard gate: it re-fetches every cited source and checks the claimed evidence against the page before a claim is allowed into the report. A bot-walled source with no archive copy is marked snippet-only, not verified.
  • Made CI run the real validation, scoring and verification code on fixture reports with saved pages, so a code change can turn the eval red; three planted bugs each did (signal-scout#3).
  • Shipped as a standalone Claude Code / OpenCode plugin rather than a hosted service, so it runs against a user's own Claude Code session with no separate backend to operate.

Outcome

cited began as this verification step and grew into its own tool. Fixes made there later came back: the check that numbers and names in a claim are on the page, and the fetcher that refuses private addresses, were ported from cited in 1.7.0.

Key features

  • Turns a startup URL into a ranked shortlist of prospects, segments, and companies
  • Every prospect cites a public source; verify_sources.py re-fetches it, a prospect whose cited evidence is clearly missing from the page is dropped before the report is written, and partial matches stay flagged for review
  • Sorts every candidate into one of three types (person, segment, company), each with its own scoring and next action
  • Ships as a standalone Claude Code / OpenCode plugin

Impact

A prospect whose cited evidence is clearly missing from its source is dropped, opener included. Partial matches stay in the report, flagged for you to check.

Tech stack

Python, Claude Code Skills, OpenCode

interested in something like this?

Get in touch