signal-scout
AGENTS.md1 starsChanged 9 days agoMITView on GitHub
This repository packages signal-scout, an agent skill that turns a startup URL or product description into an evidence-backed, source-verified shortlist of first customers (Individuals), market segments (Segments), and partnership targets (Companies), using public web signals only. The skill definition is skills/signal-scout/SKILL.md — read it and follow its workflow. It is host-agnostic: it needs web search, web fetch, and a shell with Python 3.10+ (stdlib only). Tool names in the skill are generic capabilities; map them to whatever your environment provides.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repositoryOne other file it gives its agents.
- signal-scoutSkill
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
Read skills/signal-scout/references/roadmap.md first — the roadmap is frozen on new mechanisms until real outcome data exists.
- Very specific: line 21
Run the tests: cd skills/signal-scout/scripts && python3 -m pytest (or python3 test_*.py individually).
- Very specific: line 23
Keep SKILL.md compact: schemas belong in references/report-artifact.md, mechanical checks belong in finalize.py, not prose.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
signal-scout — agent instructions
This repository packages signal-scout, an agent skill that turns a startup URL or product description into an evidence-backed, source-verified shortlist of first customers (Individuals), market segments (Segments), and partnership targets (Companies), using public web signals only.
If you were asked to run signal-scout
The skill definition is skills/signal-scout/SKILL.md — read it and follow its workflow. It is host-agnostic: it needs web search, web fetch, and a shell with Python 3.10+ (stdlib only). Tool names in the skill are generic capabilities; map them to whatever your environment provides.
Key entry points:
skills/signal-scout/references/report-artifact.md— the full JSON schema for the output artifact.skills/signal-scout/references/research-framework.md— classification rules, scoring dimensions, query buckets, BD partner motions.skills/signal-scout/scripts/finalize.py analysis.json— the one-shot pipeline: validates the JSON, verifies every cited source against its live page, generates the HTML report, exportsprospects.csv, and diffs against prior runs.--validate-onlyfor the drafting loop.
Non-negotiables (see SKILL.md "Research safely"): public signals only, no scraped contact data or data brokers, no bypassing paywalls/logins, never send outreach — draft it.
If you were asked to modify this repository
- Read
skills/signal-scout/references/roadmap.mdfirst — the roadmap is frozen on new mechanisms until real outcome data exists; prefer fixes and infrastructure over features. - Scripts are Python 3.10+ stdlib only — no new dependencies.
- Run the tests:
cd skills/signal-scout/scripts && python3 -m pytest(orpython3 test_*.pyindividually). verify_sources.pyandsignal_scout_core.pyare the product's load-bearing claim (open-book source verification); be paranoid about regressions there: an inverted check would certify fabricated evidence as "Verified".- Keep
SKILL.mdcompact: schemas belong inreferences/report-artifact.md, mechanical checks belong infinalize.py, not prose.
Install targets
Claude Code reads ~/.claude/skills/signal-scout; Codex reads ~/.agents/skills/signal-scout; OpenCode reads both. node bin/install.js writes to each of those it detects a tool for (falling back to Claude Code), and ./install.sh writes both. npx signal-scout installs the published npm package (1.0.1), which is older than this repo, so it installs an earlier version of the skill, and only to ~/.agents/skills/signal-scout; --dir overrides the destination (e.g. a project-local .claude/skills/ or .agents/skills/). Claude Cowork does not load skills from ~/.claude/skills; it uses skills enabled on the claude.ai account.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

