agentleFS
corebits-iocorebits-io/deal-origination-playbook/skills/company-discovery/SKILL.md

company-discovery

Skill0 starsChanged 2 months agoMITView on GitHub

Enumerate private companies that match an acquisition mandate, from public sources — association directories, trade show exhibitor lists, registries, licence rosters, certification registries, review platforms, job boards, sponsor portfolio pages. Records a source for every row and never fabricates a field. Triggers on: find companies, find acquisition targets, build a target list, source companies, company discovery, enumerate the market, find businesses to acquire, who could we buy, expand the target list, find off-market companies.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. kkrlstrm/gtm-pipeline★ 16same 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 repository6 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 7 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read mandate-profile.md. If it does not exist: stop and create it from templates/mandate-profile.md.

    line 21
  2. Very specific:

    Read the market map for this sector.

    line 24
  3. Distinctive:

    Ask for a target list size.

    line 49
  4. Distinctive:

    Check each source's terms before working it, and record any source whose terms prohibit collection in the coverage log rather than working …

    line 77
  5. Very specific:

    Not found is correct and expected: Use it constantly. Never fill a gap with a plausible guess

    line 95
  6. Distinctive:

    A thin website is normal for a good private company: Not a negative signal. Do not skip a company for looking small online

    line 99
  7. Distinctive:

    Do not skip a company for looking small online

    line 100
  8. Distinctive:

    Quote, do not summarize: If you cannot quote it, you do not have it

    line 101
  9. Distinctive:

    Every row carries its source URL: No exceptions

    line 102
  10. Distinctive:

    Respect robots and terms: Read pages. Do not bulk harvest sites that prohibit it

    line 107
  11. Very specific:

    Before a row is kept, check whether the name is a franchise brand.

    line 112
  12. Very specific:

    All owner and contact fields → owner-identification

    line 140

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Company Discovery

Enumerates the mapped population. The quality of this step sets the ceiling on everything downstream, so it is slow and it is sourced.

Read first:

  • references/sourcing-sources.md — where private companies are findable
  • references/ownership-taxonomy.md — what you are looking for, and the traps
  • references/scoring-rules.md — the anti-fabrication rules
  • references/research-ethics.md — terms of use, personal data

Step 0 — Load the mandate and the map

  1. Read mandate-profile.md. If it does not exist: stop and create it from templates/mandate-profile.md. A list built without a mandate is a list for somebody else's fund
  2. Read the market map for this sector. If there is none: offer to run market-map first. Say plainly why: without a map you do not know the segment boundaries, the discovery surface, or which sponsors to screen against, and the list will have to be rebuilt

🚧 HARD GATE — intermediary check

Section 9 of the mandate profile. Principal buyer: proceed. Fee-based intermediary: STOP, per references/research-ethics.md. TBD or blank: ask.


Step 1 — Confirm the brief

Restate back, from the mandate and the map:

  • Sectors and segments in scope, in the map's segment language
  • The size box: revenue, EBITDA, check/EV
  • Geography, or No constraint
  • Ownership types that are actionable
  • Explicit exclusions, portfolio conflicts, and who counts as a competitor (ask directly if this is vague — a competing sponsor's platform and a conflict with your own portfolio are different things and both matter)
  • The named sponsors from the map, for the screen

Ask for a target list size. Recommend 100 to 300 for a first run. If the user asks for thousands, say plainly why that is worse, then do what they decide.

Wait for confirmation.


Step 2 — Plan the search, then show the plan

Build the plan from the map's discovery surface, weighted to Tier 1 in references/sourcing-sources.md.

A plan for ~200 companies looks roughly like:

AngleMethodRough yield
Association directoriesMember lists, 2 to 4 bodies, plus regional chapters40 to 80
Trade show exhibitorsExhibitor lists, 2 to 3 events, several years30 to 60
Registries and licencesState registries, Companies House, licence rosters20 to 50
Certification registriesISO / sector accreditations10 to 30
Review platforms and mapsDensity sweep per metro for local services20 to 50
Local press listsBusiness journal fastest-growing and family lists10 to 30
Job boardsActive hiring as an operating signal10 to 30
Sponsor portfolio pagesWho is gone; the independents around them10 to 20

Run several angles. Each is blind to what the others surface. Good private companies are frequently visible in exactly one place.

Check each source's terms before working it, and record any source whose terms prohibit collection in the coverage log rather than working it.

Show the plan. Get a nod. Then go.


Step 3 — Research

One angle at a time. For each candidate, capture only what you can see:

  • Company name, domain, HQ location
  • Sector and segment, in the map's language
  • Year founded and employee count, if published
  • Where you found it

Rules while researching:

  1. Not found is correct and expected. Use it constantly. Never fill a gap with a plausible guess
  2. Never infer revenue or EBITDA from headcount and present it as fact. If you estimate, label it and state the basis in the same field
  3. A thin website is normal for a good private company. Not a negative signal. Do not skip a company for looking small online
  4. Quote, do not summarize. If you cannot quote it, you do not have it
  5. Every row carries its source URL. No exceptions
  6. Check the domain before merging near-identical names. Similar names are usually different companies
  7. Dead or parked site: record that as the finding. Do not describe the company from memory
  8. Respect robots and terms. Read pages. Do not bulk harvest sites that prohibit it

The franchise and brand screen

Before a row is kept, check whether the name is a franchise brand. Franchise locations look exactly like independent local businesses in every Tier 2 source. Mark them and let ownership-classification handle the classification.

Deduplicate as you go

Match on domain, not name. Keep the row with the better source. Record where the duplicate came from — appearing in three independent sources is itself a mild positive signal, and it belongs in notes.


Step 4 — Write targets.csv

Append to the workspace targets.csv, creating it from templates/targets.csv if absent.

At this stage you fill: company, domain, hq_location, sector, segment, year_founded, employee_count, source, date_added. Set status to identified.

Fill est_revenue and est_ebitda only if published, or as a labeled estimate with the basis stated. Otherwise Not found.

Leave blank for the next steps:

  • ownership_type, ownership_evidence → ownership-classification
  • fit_score, fit_reason, excluded_by → mandate-fit-scoring
  • signal, signal_source, evidence_date, signal_strength → business-research
  • All owner and contact fields → owner-identification

Everything unknown is the literal string Not found. Never blank, never a guess.

🚧 HARD GATE — before writing

Show the user:

  • How many rows you are about to add
  • How many are new vs. duplicates of existing rows
  • The breakdown by segment
  • How many are suspected franchise or branded-network locations
  • Five sample rows in full, so quality can be inspected
  • Any angle that produced little or nothing, and why

Get approval. Then write.


Step 5 — Report honestly

Discovery run complete

Added:            164 new companies
Duplicates:       19 (already in list)
Total list size:  164

By segment:
  Commercial contract-recurring    88
  Commercial project                41
  Mixed                             22
  Unclear                           13

Evidence quality:
  Founding year sourced            121
  Employee count published          64
  Revenue published                  7
  Revenue Not found                141   (16 labeled estimates)

Flags:
  Suspected franchise locations       9  → ownership-classification
  Dead or parked domain               4

Angles run:      6 of 8
Not run:         Import/export records (sector is services), job boards (deferred)
Weakest angle:   Local press lists (6 rows — city has no book-of-lists feature)
Blocked:         [Directory X] — terms prohibit collection, logged, not worked

Next: ownership-classification on all 164.

Report what did not work. A silent shortfall reads as full coverage, and that is the one failure the user cannot detect on their own. If you capped or truncated anything, say exactly where.


Notes

  • Stop and check in every 50 rows on a large run. Show a sample. Correcting the brief at row 50 is much cheaper than at row 300
  • Volume is not the goal. 150 well-sourced rows beat 1,500 scraped ones on every metric that matters
  • If yield is low, the brief is probably too narrow. Come back with what you are seeing rather than padding the list to hit a number
  • Never score or classify ownership here. Both have their own skill and their own rules, and doing them inline skips the evidence hierarchy and the rubric
  • Update the coverage log at the end of every run, per coverage-tracking

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.