Gtm Call Stack

• Source: Built by Jonathan Kvarfordt on the GTM AI OS. Supplied as a complete zip and served unchanged.
• Frameworks applied: Hub and setup

It covers a job one of the eight revenue seats has to finish, and it states its process, decision rules, and output format clearly enough for an agent to follow.

1. 1.Copy the SKILL.md text below, or download the raw file.
2. 2.Create a folder named exactly gtm-call-stack in your agent's skills directory.
3. 3.Save the file inside that folder as SKILL.md.
4. 4.Ask the agent one of the trigger requests below.
5. 5.Check the output against what you already know before it leaves your desk.

• process my calls
• tear down this transcript
• what should I do with my Granola notes

---
name: gtm-call-stack
description: Hub for the Granola-to-Claude GTM call stack. Use when someone says "process my calls", "tear down this transcript", "what should I do with my Granola notes", "run the call stack", or is unsure which call skill to use. Routes a transcript or Granola note through capture, denoise, role extraction, folder Recipes, win/loss, and the Account Context.
---

# GTM Call Stack (hub)

Transcripts capture everything the buyer said and nothing about what the rep noticed. This stack turns a noisy call into data Claude can act on, in the order of the GTM AI OS loop: KNOW, DECIDE, ACT, LEARN.

AI proposes. Evidence supports. Humans govern. Outcomes teach.

## The stack

| Step | Skill | Input | Output | Loop stage |
|---|---|---|---|---|
| 0 | `gtm-context-setup` | 3 questions | `gtm-context.md` (ICP, personas, signal library) | Setup |
| 1 | `granola-capture-markers` | Granola typed notes | Rep Read: pain, moment, fit, confidence, doubt the rep typed live | KNOW (human layer) |
| 2 | `transcript-denoise` | raw transcript | Evidence Ledger: atomic, quoted, tagged facts; noise report | KNOW |
| 3 | Role scans (see below) | Evidence Ledger | role-specific findings | KNOW, DECIDE |
| 4 | `folder-recipe-run` | a Granola Folder | one table, one row per call | KNOW at scale |
| 5 | `win-loss-icp-update` | Won and Lost folders | pattern report, ICP diff, signal weight diff | LEARN |
| 6 | `account-context-builder` | one Granola meeting via MCP | scored Account Context: FIT, PAIN, MOMENT, CONFIDENCE, next step | DECIDE, ACT |

Role scans read the Evidence Ledger. Exception: `call-coaching-scan` also reads the raw transcript for seller lines.

| Role | Skill |
|---|---|
| AE, sales leader | `deal-qualification-scan` |
| Any rep, manager, enablement | `call-coaching-scan` |
| CSM, AM | `customer-health-scan` |
| PMM, marketing, product, competitive | `voice-of-customer-scan` (includes competitor mentions) |
| AE, SDR, CSM | `follow-up-builder` |

## Routing

1. If `gtm-context.md` does not exist in the working folder, run `gtm-context-setup` first. Every scoring step depends on it.
2. Find the input:
   - Granola MCP connected (tools `list_meetings`, `get_meetings`, `get_meeting_transcript`, `list_meeting_folders`, `query_granola_meetings`): ask which meeting or folder, then fetch. `get_meetings` returns the typed notes and summary. `get_meeting_transcript` returns the verbatim transcript (paid Granola plan).
   - Pasted transcript or a `.txt` / `.md` file: use it.
   - Neither: say so and ask for one. Do not invent a call.
3. Always run Steps 1 and 2 before anything else on a single call. The typed notes run first because they are the rep's judgment and the denoise step uses them to attribute speakers.
4. Then route by the ask:
   - "Score this account", "build the account record", "what's the next step": `account-context-builder`.
   - "How did I do", "coach me": `call-coaching-scan`.
   - "Is this deal real", "MEDDICC": `deal-qualification-scan`.
   - "Is this customer at risk", "renewal": `customer-health-scan`.
   - "What are buyers saying", "messaging", "feature requests", "competitor mentions": `voice-of-customer-scan`.
   - "Write the follow-up", "recap", "CRM notes": `follow-up-builder`.
   - "Across all my calls", "this folder", "every call last month": `folder-recipe-run`.
   - "Why do we win", "why do we lose", "update the ICP": `win-loss-icp-update`.
5. If the ask fits more than one, run them in table order and say which ran.

## Rules for every skill in the stack

- Quote, do not paraphrase, anything labeled PAIN. If you cannot quote it, it is not PAIN.
- Every fact carries a source: meeting title, date, timestamp or line, speaker.
- Every fact carries a confidence tag: Verified, Supported, Inferred, Unknown.
- Rep typed notes are first-party judgment. They set the rep's confidence. They never become a buyer quote unless the same words appear in the transcript.
- Never write to a CRM, send a message, or enroll anyone. Output is for a human to review.
- No em-dashes, no filler, no jargon in output. Short lines. Tables where a reader will scan.

## Files the stack reads and writes

```
gtm-context.md            company, ICP, personas, signal library (setup)
calls/<date>-<slug>/      ledger.md, rep-read.md, scans per call
accounts/<slug>.md        the Account Context, one per account, updated in place
folders/<folder>/         folder-run table and win/loss outputs
```

Create folders as needed. Never overwrite an Account Context: update fields and append to its change log.

What does the Gtm Call Stack skill do?

Hub for the Granola-to-Claude GTM call stack.

Where does the Gtm Call Stack skill come from?

Built by Jonathan Kvarfordt on the GTM AI OS. Supplied as a complete zip and served unchanged.. The source and included files are shown on this page and in the package download.

Why was the Gtm Call Stack skill chosen for this library?

It covers a job one of the eight revenue seats has to finish, and it states its process, decision rules, and output format clearly enough for an agent to follow.

When should the Gtm Call Stack skill not be used?

Use it only for the job named above. A skill applied outside its scope produces confident output on the wrong question.

How do I install the Gtm Call Stack SKILL.md file?

Download the file, create a folder named exactly gtm-call-stack inside your agent's skills directory, and save the file inside it as SKILL.md. The agent loads it when a request matches the description.