Artemis Trajectory
The ICP Definition & Validation consulting agent — guides your diagnostic, build & support
A Claude-orchestrated implementation package that walks you through pulling win patterns from your closed-won data, mining discovery transcripts and customer interviews for the real pains, and translating both into Ideal Customer Problems plus a tuned ICP scorecard wired into your CRM. Includes quarterly, monthly, and annual refresh routines so the ICP never goes stale.
What is Artemis Trajectory?
Replaces firmographic-only ICP with a Pain-First framework validated against closed-won deals. The Artemis ICP Definition skill builds an ideal customer profile from win patterns and the acute pains behind your closed-won deals, not just firmographics. The output is a set of Ideal Customer Problems, a 5-7 dimension ICP scorecard tuned so most of your recent wins land in the Ideal band, A/B/C/D tier definitions, a disqualification list of firmographically-right-but-failing accounts, CRM segment setup, and quarterly, monthly, and annual refresh routines.
What this solves
Wrong ICP / Poor Targeting
Replace firmographic-only ICP with the Pain-First framework
Assumed stack (adapts to yours)
CRM + customer research transcripts (Granola, Sybill, or Attention)
This is a consulting engagement that happens to cost $349
When you run this skill in Claude, you don't get a playbook execution. You get a senior GTM advisor walking you through the build — diagnosing your stack one question at a time, branching to your reality, pasting screenshots back and forth, naming gotchas before you hit them, and surfacing the other things in your stack that need automation along the way.
Discovery-first
Claude opens with discovery — one question at a time — before recommending a single tool. Your stack, stage, and constraints drive every downstream call.
Walkthrough, not handoff
Paste your CRM config back. Take a screenshot of your routing rule. Run the query and tell Claude what came back. Every phase has a paste-back moment.
Running observations
Claude notices the other broken or manual systems in your stack as you build and surfaces them at session end — a "what I'd automate next" list, prioritized.
How the engagement runs
Claude walks you through these phases — discovery, build, verification, closing summary — the same way a senior GTM advisor would on a $40K consulting retainer.
- 1
Discovery
Read your current ICP, closed-won and closed-lost inventory, and conversation-intel coverage. The build branches off how much real data you have to work with.
- 2
Win Pattern Analysis
Tier your closed-wons, then cluster the firmographic, trigger, and buying-committee patterns that show up in your best customers but not your lost deals.
- 3
Pain Extraction
Mine 10-20 discovery transcripts and run 3-5 best-customer interviews to surface the acute pains behind each win. This is the unfakeable part.
- 4
Ideal Customer Problems + Scorecard
Write 3-5 problem statements, then build a 5-7 dimension scorecard tuned so 8 of your last 10 wins land in the Ideal band.
- 5
CRM Segment Setup
Create the scoring properties, automation, and routing rules in HubSpot, Salesforce, or Attio so the ICP is operational, not a slide nobody opens.
- 6
Install Routines
Wire quarterly refresh, monthly fit-distribution, and annual deep-refresh routines so win-pattern drift gets caught before the ICP becomes folklore.
What's in the bundle
Dual-format. Same playbook, your choice of interface.
For Claude Chat
Pre-built Claude Project. Paste the system prompt, upload supporting files, send the opening message. No CLI required.
- Claude Project system prompt
- Opening message to kick off discovery
- Template files to upload (CRM workflows, sequences, configs)
- Step-by-step install guide for claude.ai
For Claude Code
Full skill bundle with sub-agents and ongoing routines. Wires the system into your local environment for long-term operation.
- skill.md (drops into ~/.claude/skills/)
- Specialized sub-agents (CRM mapper, sequence writer)
- Routines folder — weekly + monthly + quarterly checks
- CronCreate-ready scheduled prompts
- All template files used by the Chat variant
What's actually in the .zip
You're buying a downloadable bundle, not a service. Here's the exact file structure.
icp-definition-skill-v1/ ├── README.md # Buyer install picker (Chat vs Code) ├── shared/ │ ├── playbook.md # Source-of-truth implementation guide │ ├── partner-stack.md # Artemis partner stack positioning │ └── templates/ │ ├── README.md # Template inventory + scaffolding │ └── (skill-specific templates) ├── claude-chat/ │ └── system-prompt.md # Paste into a Claude Project ├── claude-code/ │ ├── SKILL.md # Drops into ~/.claude/skills/icp-definition/ │ ├── agents/ │ │ └── (2-3 sub-agent .md files) │ └── routines/ │ └── (3 CronCreate-ready routine .md files) └── LICENSE.txt # Single-user commercial license
Format
.zip · markdown
License
Single-user commercial
Updates
1 refresh included
How partner recommendations work
This Skill Package recommends tools from Artemis GTM partner companies (Warmly, Amplemarket, Maildoso, Attio, Sybill, Attention, Apollo, RB2B — depending on the skill). When you click through and sign up via the in-skill referral links, Artemis GTM may earn a commission, at no extra cost to you.
We recommend these tools based on validation across 127+ B2B SaaS audits, not because they pay us. The skill is built to adapt to your existing stack — if a tool you have is working, you keep it.
For current partner pricing and feature detail, the skill defers to each partner's website. Claude is explicitly instructed not to invent specifics — when a buyer asks "how much does Tool X cost?", Claude refers to the partner's homepage rather than quoting a potentially stale number.
Built on the same playbook our consulting uses
Read the source content this skill is built from. Everything in the bundle is the operational version of these guides.
Frequently Asked Questions
Ready to ship wrong icp / poor targeting?
One-time purchase — yours to keep, with one update refresh. Always-latest updates come with Artemis membership.