CLAUDE.md template (revenue team)

Working CLAUDE.md template for a revenue-team Claude Code repo. Eight sections covering identity, ICP, voice, signal scoring, data contract, escalation rules, few-shot examples, and audit metrics. Replace bracketed values with your team's data; keep the section structure.

Save as CLAUDE.md at the root of your GTM Claude Code repo. Every subagent reads it before drafting.

1. Identity and audience

Open with two paragraphs naming who the team is, what the company sells, and who's reading the output. The agent uses this to calibrate everything that follows.

We're the GTM engineering team at [Company Name]. We sell [product] to [primary buyer persona] at [company segment]. The recipients of our outbound are senior operators who get 200+ cold messages a month. They unsubscribe from anything that reads as automated. The bar for our messages is whether the recipient would forward this to their boss, peer, or team.

2. ICP definition

• Company segment. [Specific industry, size band, funding stage, geography]
• Buyer titles. Primary + secondary, with seniority floor.
• Triggering signals. 5-7 specific events that justify outbound.
• Required STA threshold. 7/9.
• Exclusions. Lookalike segments to skip even if they look qualified.

3. Voice rules and banned phrases

• Tone. Direct, operator-level, no hedging. First person plural.
• Banned punctuation. Em dashes. Use periods, commas, or rewrite.
• Banned phrases. "Just checking in." "Game-changer." "Robust." "Leverage" as verb. "Not just X, it's Y." Plus your team's specific list.
• Banned structures. Negative parallelism. Three-fragment dramatic triads. Self-answered rhetorical questions. Filler intensifiers.
• Required structures. Trigger first, problem second, concrete offer third, value-prop last.

4. Signal scoring rubric (STA)

Score every draft 0-3 per criterion. Pass: 7/9. Three failed regens, skip. See the full TVA rubric for the scoring detail.

5. Data contract

The data sources the agents are allowed to read and the fields valid for personalization.

• CRM exports. Path, fields, refresh cadence. Read-only.
• Signal sources. Hiring (BuiltIn, LinkedIn jobs API), funding (Crunchbase), leadership (LinkedIn), tech stack (BuiltWith), reviews (G2, Capterra), regulatory (SEC EDGAR, state SOS).
• Contact data. Validated within last 60 days. Title fields cross-checked against company website.
• Hallucination risks (banned citations). Unverified titles, unverified salary, quotes without URLs, internal strategy not publicly stated.

6. Escalation rules

• Three failed regenerations on the same prospect, skip and log.
• Signal references a person by name, require human approval.
• Signal touches regulated topics (PHI, MNPI, FERPA), require human approval.
• First 50 sends per new ICP segment, require human approval.
• Any send that would cite an unverified data point: regenerate without the citation.

7. Few-shot examples

The leverage of the whole CLAUDE.md is here. Five excellent messages from your team, each annotated with the trigger, STA score, the final message verbatim, and the outcome. Update monthly.

8. Audit and metric definitions

Every draft logs prospect, signal source URL, STA score, regeneration count, final decision. Every send logs recipient, send time, message hash, audit trail.

Optimization metric: forward rate. Reply rate is secondary. Open rate is not used for decisions.

Maintenance cadence

Weekly for the first month. Monthly after that. Triggers to update: a forwardable send (add to few-shot), a polite decline (refine banned phrases), a winning signal (raise weight), a wasted signal (drop from qualifying list).

A frozen CLAUDE.md is a dead system.

Where this fits

The CLAUDE.md is the contract. The TVA scoring rubric is the gate. The ICP rebuild worksheet populates section 2. Together they're the starting kit for a GTM Claude Code repo. See the architecture overview here.