icp-scoring
**Duration:** 15–30 minutes per account (or run in batch) **Output:** ICP score + tier assignment saved to CRM or…
**Duration:** 20–40 minutes per account **Output:** `outputs/[date]-[account]-research.md`
$ npx -y skills add karlraf/gtm-starter-kit --skill account-research --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/account-researchContext preview
The summary Claude sees to decide when to auto-load this skill.
**Duration:** 20–40 minutes per account **Output:** `outputs/[date]-[account]-research.md`
**Duration:** 20–40 minutes per account **Output:** `outputs/[date]-[account]-research.md`
---
Read skills/account-research/SKILL.md and research [company.com]
That's it. Claude will read your ICP, signals, and competitor context automatically and produce a structured research brief. See `examples/sample-company/outputs/2024-03-15-research-northstar-analytics.md` for what the output looks like.
---
Build a complete intelligence brief on a target account before any outreach. The goal is not a company summary you could get from their About page. The goal is to find the specific trigger — the moment, the hire, the product shift, the funding event — that makes this the right time to reach out, and the right angle to use.
---
---
---
Pull the basics. Don't stop here — this is table stakes.
Sources: LinkedIn, Crunchbase, BuiltWith, their own blog and changelog.
---
Identify the 2–3 people to reach. Reference `context/personas/` for titles and roles to target.
For each stakeholder:
---
Reference `context/signal-library.md`. For each Tier 1 and Tier 2 signal:
If score ≥ 60: outreach should go out within 48 hours. If score < 40: note the gap and flag what would change the picture.
---
Reference `context/competitor-radar.md`.
---
This is the only thing that matters. Everything above was research. This step is judgment.
Answer these questions:
1. **Why now?** What's the specific event or condition that makes this the right moment to reach out? If you can't answer this, don't reach out yet.
2. **Why us?** What's the specific capability or outcome that maps to their current situation — not our general value prop, but the specific thing they need right now?
3. **What's the hook?** The first line of the email. It should reference something specific — a recent hire, a product move, a funding event — and connect it to a relevant insight. Not "I saw you raised Series B." Something they'd actually want to read.
4. **Who leads?** Which stakeholder gets the first touch, and in what channel?
---
# Account Research: [Company Name] Date: [YYYY-MM-DD] Researched by: [Claude / Name] Signal score: [X/100] Recommended action: [Immediate outreach / Sequence / Monitor / Skip] ## Company Snapshot [2–3 sentences: what they do, stage, recent momentum] ## Funding & Growth - Last round: [Amount, Date, Lead] - Headcount: [X, growth rate] - Key recent hires: [Roles, dates] ## Tech Stack - [Tools identified] - Integration opportunities: [List] - Displacement signals: [List] ## Stakeholder Map | Name | Title | Tenure | Reach via | Notes | |------|-------|--------|-----------|-------| | [Name] | [Title] | [X months] | [Channel] | [Key context] | ## Active Signals | Signal | Status | Fired | Score contribution | |--------|--------|-------|-------------------| | [Signal] | Active/Inactive | [Date] | +[X] pts | ## Competitive Context [Any competitor usage or evaluation signals] ## The Angle **Why now:** [Specific trigger] **Why us:** [Specific fit] **Hook:** [Proposed first line] **Recommended sender:** [Name / Role] ## Suggested Next Action [Specific: which sequence, which template, which stakeholder, which channel]
---
Before filing the output, confirm:
Built by The Revenue Architects — a GTM engineering firm based in San Francisco.
**Duration:** 15–30 minutes per account (or run in batch) **Output:** ICP score + tier assignment saved to CRM or…
**Duration:** 15–30 minutes (including your review) **Run once:** When you first clone this repo **Output:** All context files…
**Duration:** 2–4 hours per campaign **Output:** `outputs/campaigns/[date]-[campaign-name]/`
**Duration:** 10–15 minutes **Cadence:** Weekly (Monday morning recommended) **Output:** Updated context files + a brief change…