business-research
Skill0 starsChanged 3 months agoMITView on GitHub
Research a target business using public web sources and produce structured, evidence-graded findings — verified facts, inference, and confidence kept separate. Use when researching the locked LordGen demo business or a candidate business.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- tony/skillstheir own version, first seen Oct 31, 2026
- corebits-io/deal-origination-playbooktheir own version, first seen Oct 13, 2026
- tarrragon/claude-skillstheir own version, first seen Sep 1, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- tarrragon/claude-skills★ 1same name
- corebits-io/deal-origination-playbook★ 0same name
- tony/skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- automation-builderSkill
- competition-demoSkill
- execution-planSkill
- lordgen-3d-web-experienceSkill
- lordgen-full-rebuildSkill
- lordgen-site-upgradeSkill
- opportunity-scoreSkill
- proposal-generatorSkill
- workflow-qaSkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 2 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Use Claude Code's built-in WebSearch and WebFetch tools only — no external research MCP server (per the approved architecture, …
- Very specific: line 22
If firecrawl also fails (e.g.
- Distinctive: line 23
Never fabricate page content.
- Very specific: line 29
Write data/research.json with this structure:
- Very specific: line 50
Every fact and pain_point must carry a source_url (or "source_url": null with an explanation if genuinely unsourceable) and a type of …
- Distinctive: line 52
If a source is unreachable, record that as a data caveat rather than silently dropping it or guessing at its content.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Business Research
Purpose
Research a target business and produce structured findings that a human can trust: what's actually known, what's a reasoned guess, and how confident each claim is. Never present inference as verified fact (CLAUDE.md §11).
Method
Use Claude Code's built-in WebSearch and WebFetch tools only — no external research MCP server (per the approved architecture, docs/architecture.md). Prioritize, in order:
- The business's own website — services, hours, service area, stated policies/claims.
- Regulatory/accreditation sources (e.g. BBB) — rating, accreditation status, complaint patterns.
- Independent review aggregators (Google, Yelp, Trustpilot, industry-specific sites) — ratings, review counts, recurring themes.
Some sources (Yelp, Trustpilot, and similar) commonly block automated WebFetch access (HTTP 403). When that happens, in order:
- Retry with
firecrawl_scrape(proxy: "stealth"or"enhanced") — approved 2026-08-13 specifically for this fallback case (docs/architecture.md). This gets past most bot-blocks and returns primary-source content, not a summary. - If firecrawl also fails (e.g. still 403, or the content is behind client-side JS pagination that doesn't render), fall back to
WebSearchresult summaries as a last resort. - Never fabricate page content.
Mark facts sourced via WebSearch summary (rather than a direct page read, by either tool) with "confidence": "low" and a "caveat" explaining why. Facts read directly (via WebFetch or firecrawl_scrape) can carry "confidence": "high" for the fact's existence/text — but a single review/account being real and directly-read is not the same as it being representative; that judgment still belongs in confidence on the pain_points/automation_opportunities entries that cite it, not in inflating the underlying fact.
Required Output
Write data/research.json with this structure:
{
"business": { "name": "", "category": "", "founded": "", "address": "", "phone": "", "service_area": "", "hours": "" },
"research_date": "YYYY-MM-DD",
"facts": [
{ "fact": "", "type": "verified|inference", "source": "", "source_url": "", "confidence": "high|medium|low", "caveat": "" }
],
"pain_points": [
{ "description": "", "type": "verified|inference", "evidence": "", "confidence": "high|medium|low", "supporting_fact_refs": [] }
],
"automation_opportunities": [
{ "opportunity": "", "rationale": "", "confidence": "high|medium|low", "supporting_fact_refs": [] }
],
"data_caveats": [""],
"overall_confidence": "high|medium|low"
}Rules:
- Every
factandpain_pointmust carry asource_url(or"source_url": nullwith an explanation if genuinely unsourceable) and atypeofverifiedorinference— never omittype. automation_opportunitiesare candidate opportunities only. Scoring/ranking them is theopportunity-scoreskill's job (Phase 3), not this one.- If a source is unreachable, record that as a data caveat rather than silently dropping it or guessing at its content.
overall_confidencereflects the weakest link, not the average — e.g. if the core pain-point evidence is only "low," overall confidence should not read "high."
Out of Scope
This skill only researches and writes data/research.json. It does not score opportunities, write proposals, or take any external action.
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.

