---
name: gtm-context-setup
description: Build the gtm-context.md file every call skill reads: what you sell, ICP band and gates, personas, competitors, proof points, and the MOMENT signal library. Use when someone says "set up my GTM context", "set up the call stack", when gtm-context.md is missing, or when a call skill needs the ICP to score FIT.
---

# GTM Context Setup

Context before automation. A transcript scored against no ICP produces confident noise. This skill writes one file, `gtm-context.md`, in the working folder. Every other skill in the stack reads it.

## Step 1: Pick a mode

Ask one question: "Fast (3 questions, about 3 minutes) or in-depth (12 questions, about 15 minutes)?" Default to fast if they do not answer.

If `gtm-context.md` already exists, read it, show what it holds, and ask only for what is missing.

## Step 2: Fast mode, 3 questions

Ask one at a time. Take their words. Do not polish them.

1. **What do you sell, and who buys it?** One or two sentences. Get the product, the problem it solves, and the buyer title.
2. **Your ICP band.** Employee range (floor and ceiling), top 1 to 3 industries, regions. Then: "Which of those must be true or the account does not count?" Those are the gates.
3. **Who is the champion and who signs?** Champion titles and economic buyer titles.

Fill the rest with the defaults below and label each one `DEFAULT, replace with yours` in the file.

## Step 3: In-depth mode adds

4. Tools you replace (replacement targets).
5. Top 3 competitors and how buyers name them.
6. FIT weights per ICP dimension (otherwise equal weights, labeled default).
7. High FIT cutoff on a 0 to 10 scale.
8. MOMENT window in days.
9. Your own signal weights if you have win/loss data.
10. Proof points: 2 or 3 claims with a number and a named or anonymized customer.
11. Disqualifiers: things that, said on a call, end the deal.
12. Where your call notes and CRM live (so no skill assumes a system).

## Defaults (fast mode only, always labeled)

| Setting | Default |
|---|---|
| FIT weights | equal across stated dimensions |
| High FIT cutoff | 7.0 |
| MOMENT window | 90 days |
| Signal library | public or call pain quote 10, competitor dissatisfaction 8, new executive 6, job posting 5, tech installed 4, funding 3, engagement 3, deadline or event stated on a call 6, budget allocated stated on a call 5 |
| Proof points | none. Value and Proof lines in drafts stay as marked gaps |

The signal library is the GTM AI OS starting library plus two call-native signals. It has not been fit to anyone's close rates. `win-loss-icp-update` replaces it with evidence.

## Step 4: Write the file

Use exactly this structure so other skills can parse it:

```markdown
# GTM Context
Updated: YYYY-MM-DD | Mode: fast | in-depth

## What we sell
## Buyer problem we solve
## ICP band
| Dimension | Value | Gate (Y/N) | Weight |
## High FIT cutoff
## Personas
| Role | Titles |
| Champion | |
| Economic buyer | |
| Influencers | |
## Replacement targets
## Competitors
| Name | How buyers say it |
## Disqualifiers
## Proof points
## MOMENT window (days)
## Signal library
| Signal | Points | Source (default or win/loss date) |
## Where data lives
## Change log
```

Show the file. Say which values are defaults. Offer one line: "Replace the defaults as you learn. Run win-loss-icp-update when you have 10 or more closed calls."