business-research
Skill0 starsChanged 2 months agoMITView on GitHub
Research each target company properly before anyone speaks to it — what the business does, scale evidence, tenure and history, succession and ownership-change signals quoted and dated, what the owner appears to care about, and an explicit list of what is not known. Produces a one-page target brief per company. Triggers on: research this company, research the owners, target brief, brief me on, pre-call research, succession signals, deal brief, prepare for the call, what do we know about this company, read the signals.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- tony/skillstheir own version, first seen Oct 31, 2026
- Kikobazz123/lordgen-competition-demotheir own version, first seen Oct 11, 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
- Kikobazz123/lordgen-competition-demo★ 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 repository6 other files it gives its agents.
- company-discoverySkill
- coverage-trackingSkill
- mandate-fit-scoringSkill
- market-mapSkill
- owner-identificationSkill
- ownership-classificationSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
15 rules. 0 are common, 9 are distinctive, and 6 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
9
Distinctive
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 23
Read mandate-profile.md — section 2 for the thesis, section 10 for what the user is cleared to say about themselves
- Distinctive: line 25
Read the market map — the brief must describe the company in the map's segment language
- Very specific: line 30
Never brief a
Not a FitorNear Missrow: If the user asks for one specifically, do it and say why the row is not in the keep pile. - Distinctive: line 33
Ask how many to brief. Recommend 10 to 20 at a time.
- Very specific: line 89
All three, or it is Not found.
- Distinctive: line 96
Before recording any signal, check it against this list.
- Very specific: line 117
Use templates/target-brief.md, one file per company, into briefs/ in the workspace.
- Distinctive: line 120
Three specifics only someone who read about them would know: Not flattery, not generic sector observations.
- Distinctive: line 123
Not a discovery checklist
- Distinctive: line 131
Keep the brief to one page of substance.
- Very specific: line 141
Update est_revenue / est_ebitda only with a published figure or a labeled estimate with the basis stated inline.
- Very specific: line 144
If research contradicted the earlier classification — sponsor ownership found, family rather than founder, a different segment — say so …
- Distinctive: line 155
Every company where you found no signal at all
- Distinctive: line 156
Every contradiction with the existing row
- Distinctive: line 217
This is where the playbook stops: No outreach copy, no sequences, nothing sent
“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
Produces the one page you take into a first conversation: what this company is, what the evidence says, what the signal is, and — the part that matters most — what you do not know.
Read first:
references/signals-and-evidence.md— signal classes, the evidence standard, what is noise, the line you do not crossreferences/ownership-taxonomy.md— ownership context and succession dynamicsreferences/research-ethics.md— public professional sources onlyreferences/scoring-rules.md— anti-fabrication rules
Step 0 — Load and scope
- Read
mandate-profile.md— section 2 for the thesis, section 10 for what the user is cleared to say about themselves - Read the market map — the brief must describe the company in the map's segment language
- Read
targets.csv - Filter to
fit_score=Strong FitorFit
Never brief a Not a Fit or Near Miss row. If the user asks for one
specifically, do it and say why the row is not in the keep pile.
Ask how many to brief. Recommend 10 to 20 at a time. A brief is 30 to 60 minutes of real research, and a rushed brief is worse than none — it produces a confident first conversation built on inference.
Step 1 — Research the business
Sources, in order of value:
- The company's own site — services, segments, locations, history, team
- Registry filings — formation, officers, changes, addresses
- Licence and certification rosters — dated proof of operating standard
- Association listings and committee roles
- Trade press and local press — anniversary pieces, award profiles, expansion coverage
- Job postings — what they are hiring tells you what they are doing
- Review platforms — volume and age, for scale and tenure proxies
- Procurement and contract award records, where public
What to establish:
| Field | Standard |
| What it does | In the map's segment language, no unsourced adjectives |
| Scale | Published figure, or labeled estimate with basis, or Not found |
| Ownership | Already classified — verify it still holds, note if it changed |
| Tenure and history | Dated events with sources |
| Locations | Opened, closed, or acquired, with dates |
| Structural changes | Name changes, new entities, holdco insertions |
Rules:
- Quote, do not paraphrase, anything you will treat as a signal
- Never infer financials from headcount and present them as fact
- A dated website is not a finding about the business. See the non-signal
table in
references/signals-and-evidence.md - Record what you checked and found nothing in. That belongs in the brief
- Date everything.
evidence_dateon every claim that has one
Step 2 — Read the signals
Work the four classes in references/signals-and-evidence.md:
- A. Succession — tenure with no visible successor, retirement language, a step back to Chairman, a second generation visibly present or visibly absent, association roles handed over
- B. Ownership change — officer changes, PSC updates, new registered agent, a holdco inserted, share transfers, name changes
- C. Operating change — new facility, certification added, hiring burst, large contract, capability added
- D. Stress — licence lapse, closure, overdue filings, litigation of record
For each signal found: the exact quote or filed fact, the URL, and the date.
All three, or it is Not found.
Score signal_strength per the table in
references/signals-and-evidence.md — high, medium, low, or Not found.
🚧 HARD GATE — the inference line
Before recording any signal, check it against this list. If it comes from any of these, do not record it and do not mention it:
- An inference about anybody's health, age, or family circumstances
- A personal social media account
- Anything obtained by pretexting — posing as a customer, supplier, applicant, or journalist
- Family members who are not in the business
- A photograph read for age or infirmity
A signal is something a person said or filed in public, in a professional context. Beyond the ethics, inferred personal signals are usually wrong, and acting on one in a first conversation is unrecoverable.
Also check independence: three articles quoting one press release are one signal, not three. Record it once and cite the primary source.
Step 3 — Write the brief
Use templates/target-brief.md, one file per company, into briefs/ in the
workspace. Every section, including:
- Three specifics only someone who read about them would know. Not flattery, not generic sector observations. Concrete facts about this company
- Questions to ask — the real ones this research raised and could not answer. Not a discovery checklist
- What we do not know — explicit, itemized, and never filled in with an inference at the last minute
The last section is the point of the whole document. A brief that hides its gaps produces a confident conversation built on an invention, and owners in small sectors remember that.
Keep the brief to one page of substance. Length is not thoroughness; a brief nobody reads before the call has failed.
Step 4 — Write back to targets.csv
Fill signal, signal_source, evidence_date, signal_strength, and set
brief_status = briefed and status = briefed.
Update est_revenue / est_ebitda only with a published figure or a labeled
estimate with the basis stated inline.
If research contradicted the earlier classification — sponsor ownership found,
family rather than founder, a different segment — say so loudly and flag the row
back to ownership-classification or mandate-fit-scoring. Do not quietly
overwrite a scored row.
🚧 HARD GATE — before writing
Show:
- Which companies you briefed
- One complete brief in full, so the user can judge the standard
- The signal found for each, with strength, quoted
- Every company where you found no signal at all
- Every contradiction with the existing row
- Anything you checked and could not establish
Get approval. Then write.
Step 5 — Report
Business research complete — 14 companies briefed
Signals found:
high 4 (2 owner-stated succession, 2 filed ownership change)
medium 6 (long tenure + no visible successor, both sourced)
low 2 (operating change only)
Not found 2 ← approach on mandate fit and merit, no angle
By class:
A succession 7
B ownership change 3
C operating 5
D stress 1 (overdue filings — diligence note, not an angle)
Evidence recency:
Under 12 months 8
1-3 years 4
Over 3 years 0 ← none presented as current intent
Rejected as non-signals: 11
(dated websites 4, no profile 3, small headcount 2, sector consolidation 2)
Rejected on the inference line: 2
(an inferred retirement age, a personal social post — neither recorded)
Contradictions found:
[Company A] — holdco inserted 2025, likely sponsor deal → back to classification
[Company B] — second generation runs ops → family-owned, not founder-owned
Scale evidence:
Revenue published 2
Labeled estimate 6
Not found 6
Re-score recommended: 10 rows now have a sourced signal and may reach Strong Fit.Notes
- Fit and signal are different questions. This skill answers the second. Never let a strong signal argue a company into the mandate
Not foundon signal is fine. It means you go on mandate fit and general merit rather than a specific angle. It does not lowerfit_score- Recency bands matter. A 2018 quote about retiring soon is a 2018 fact — and if it happened, the company may be somebody's add-on now. Check
- Re-score after a research batch. Rows blocked from Strong Fit only by a missing signal can now be reassessed
- Briefs go stale. Re-check before any second cycle
- Never commit briefs. They hold personal data and later conversation notes.
The
.gitignorecovers the workspace patterns; keep it that way - This is where the playbook stops. No outreach copy, no sequences, nothing sent
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.

