decision-maker-finder

---
name: decision-maker-finder
description: Find the economic buyer and champion at a target company, each with public source URL and still-there check, when a user needs the right names to contact.
---

# Find who signs and who will champion you

Two sourced names per account, buyer and champion, in contact order.

## Claude does
1. Ask once: company, website, what the user sells and the pain it fixes, headcount if known. No offer, no search; unknown headcount means 50-200, stated.
2. Resolve the entity: a domain is not the company; pick the parent, subsidiary or brand that would sign.
3. Define buying roles before searching: a decision-maker is a role relative to the deal, not a seniority tier. Name the pain-owning function and one line each for Economic Buyer, Champion, Technical Evaluator, End User, Blocker, Coach.
4. Committee size: under 20 staff the CEO is buyer, champion and evaluator; 50-200 staff, 3-5 people; over 200, 5-8.
5. Filter by function and level, never titles or "VP and above". Buyer: head of that function. Champion: Manager to Director in the pain department, 6-36 months in post, never C-suite.
6. Cap at 1-3 owners per role; fill placeholders, hand off.

## Then Codex does
The bulk half: company pages, templated LinkedIn and news searches, org chart, candidate table, access score, owner JSON and contact plan, into research.md, DECISION-MAKERS.md and owner.json.

```
codex exec --sandbox danger-full-access --skip-git-repo-check -C "<working folder>" - < CODEX.md
```

Fill: {COMPANY}, {COMPANY_URL}, {ENTITY}, {OFFER}, {FUNCTION}, {HEADCOUNT_BAND}, {COMMITTEE_SIZE}, {BUYER_FILTER}, {CHAMPION_FILTER}.

## Claude checks
1. owner.json: every name has source_url and still_here yes, else null; open each source_url, confirm name and title today.
2. Buyer matches {BUYER_FILTER}; no C-suite Champion; nobody picked by seniority.
3. Tenure stated: under 6 months "ramping", over 3 years "change-resistant".
4. At most 3 per role; private profiles "title only"; emails and phones only if marked *unverified*.
5. Champion day 0, buyer day 2-3 in a separate thread; no CC, no reused message.
6. Champions "untested" with evidence on power, vested interest, internal selling; helpful without power is a Coach.

Any failure: paste it atop CODEX.md, rerun until all pass. Buyer null twice: report NO PLAUSIBLE BUYER FOUND.

## Rules
- Public, professional information only; never guess personal emails, phones or addresses. Inferred emails are *unverified*; never one finder, never unvalidated sends.
- No plausible buyer? Say so; a wrong name costs more than none.
- Filter by buying role, never seniority; C-suite are buyers, not champions.
- No name without a source URL and still-there check; resolve parent and subsidiary first; 1-3 owners per role, never the whole org.
- Never single-thread, CC roles together or send one message to all.
- Helpful without power, vested interest and internal selling is a Coach, not a Champion; keep re-testing, friendliness is not evidence.
- Title is not authority: small-company managers decide, big-company VPs may need the board.
- Never ignore tenure; private profile means title-only scoring, stated.

## If Codex is not installed
Claude does the Codex half itself from CODEX.md and says so.

## Built from
- Clay, https://www.clay.com/guides/how-to-find-decision-makers-from-company-url (official guide, May 2026)
- anthropics/knowledge-work-plugins account-research, https://github.com/anthropics/knowledge-work-plugins/blob/main/sales/skills/account-research/SKILL.md (23,876 repo stars, api.github.com)
- zubair-trabzada/ai-sales-team-claude sales-contacts, https://github.com/zubair-trabzada/ai-sales-team-claude/blob/main/skills/sales-contacts/SKILL.md (1,093 stars, api.github.com)
- MEDDICC Ltd, https://meddicc.com/what-is-meddpicc/champion (framework publisher)
- louisblythe/Sales-Skills decision-maker-identification, https://github.com/louisblythe/Sales-Skills/blob/main/skills/decision-maker-identification/SKILL.md (143 stars, api.github.com)

# Find who signs and who will champion you - Codex task

No other context exists.

## You are given
- {COMPANY}, {COMPANY_URL}: target and website
- {ENTITY}: the buying parent, subsidiary or brand to research, not the bare domain
- {OFFER}: what the user sells and the pain it fixes
- {FUNCTION}: the department feeling that pain
- {HEADCOUNT_BAND}: e.g. under 20, 50-200, over 200
- {COMMITTEE_SIZE}: committee members to expect
- {BUYER_FILTER}: function and level for the Economic Buyer
- {CHAMPION_FILTER}: function, level and tenure window for the Champion

## Produce
Use curl and web search. Write three files in the working folder.

1. `research.md`: fetch {COMPANY_URL} plus /team, /about, /leadership, /press, /investors. Search "{COMPANY} news", "{COMPANY} funding", "{COMPANY} careers", "{COMPANY} Head of {FUNCTION} LinkedIn", "{COMPANY} Director {FUNCTION} LinkedIn", "{COMPANY} CEO LinkedIn", then "[person] {COMPANY} LinkedIn" per name. Log URLs with dates. List 90-day trigger events: funding, leadership change, hiring surge in {FUNCTION}, new role.

2. `DECISION-MAKERS.md`, in order:
   - Org chart for {FUNCTION} and above, sized to {COMMITTEE_SIZE}; "[Unknown - likely exists]" for unnamed roles.
   - Candidates table, at most 3 per role matching {BUYER_FILTER} or {CHAMPION_FILTER}. Columns: name, exact title, role (Economic Buyer, Champion, Technical Evaluator, End User, Blocker, Coach), LinkedIn URL, tenure (start month/year; "ramping" under 6 months, "change-resistant" over 3 years), still-there source (dated, not LinkedIn alone), 90-day signals (frustration posts, competitor engagement, reports shared, job change, or "profile private, title only").
   - Contact Access Score 0-100 from four 0-25 sub-scores: decision-makers identified, contact information, personalisation material, warm paths; one line each.
   - Contact plan, 2-4 people: Champion day 0, most personalised angle, one talking point from research.md; Economic Buyer day 2-3, separate thread, different message; others after. Ask the Champion "When decisions like this get made, who usually has the final say?"
   - Best entry point: one person and why. Sources: all URLs.

3. `owner.json`: strict JSON array, Economic Buyer then Champion: {"role": "...", "owner_name": "full name", "owner_title": "exact current title", "confidence": "high|medium|low", "still_here": "yes|no|unclear", "source_url": "the one page you trusted most"}. No clear owner on a public source: confidence "low", owner_name null.

## Rules
- Public, professional information only; never guess personal emails, phones or addresses.
- Inferred emails are *unverified*.
- No name without a source; nobody who has left.
- Stay inside the filters: no C-suite Champion, no seniority picks, no whole-org dump.
- Every Champion is "untested"; note evidence on power, vested interest, internal selling.
- Private profile: "profile private, title only", no invented signals.
- Never plan a CC or one message for two roles.

## Return
Print, in order: owner.json in full; the candidates table; Contact Access Score with sub-scores; unfilled fields and why; the three file paths.