run-all

Files (1)

• SKILL.md

SKILL.md (first lines)

# Full Pipeline Sequence

This skill runs all GTM steps in sequence. Research, qualification, and drafting can continue automatically, but every external send requires explicit user approval.

**IMPORTANT:** Read identity, ICP, voice rules, resources, plan, and storage from `${CLAUDE_PLUGIN_DATA}/config.json`. If it is missing, stop and tell the user to run `/gtm-kit:setup`.

## Preflight Checklist (MANDATORY — do this FIRST)

Complete ALL of these before running any pipeline steps.

### 1. Storage Setup

**Check if `${CLAUDE_PLUGIN_DATA}/config.json` exists in the plugin data directory.**

**If ${CLAUDE_PLUGIN_DATA}/config.json exists (Pro plan):**
- Read the API URL and key from `${CLAUDE_PLUGIN_DATA}/config.json`
- Before each major agent step, fetch latest leads, messages, connections, and needed config from the dashboard using `references/api.md`
- Data will sync to the GTM Command Center dashboard via browser tab (see `references/storage.md`)
- No spreadsheet setup needed — skip to step 2

**If ${CLAUDE_PLUGIN_DATA}/config.json does NOT exist (Basic plan):**
- Check agent memory for `sheet-config.json`. If missing, ask the user which spreadsheet platform they'd like to use:
  - **Google Sheets** — I'll use Apps Script for fast batch writes
  - **Excel Online (OneDrive/SharePoint)** — I'll use Office Scripts or direct browser editing
  - **Excel (local file)** — I'll read/write a .xlsx file on your device
- Get the URL or file path and save to agent memory as `sheet-config.json`
- Do NOT proceed without it

### 2. Session Customization

Ask the user:
> "Is there anything custom you'd like to adjust for this session (different target roles, industries, geographies, pitch angle), or should I proceed with your saved setup?"

If the user says no or wants to proceed as-is, continue immediately.

### 3. Get Browser Permissions

Navigate to each of these sites to trigger Claude's permission prompts. Do them all now so the user only has to approve once:

1. **LinkedIn** — Navigate to `https://www.linkedin.com/feed/` and confirm the page loads
2. **Spreadsheet** (Basic plan only, Google Sheets or Excel Online) — Navigate to the user's spreadsheet URL and confirm it loads

Tell the user in chat:

> "I need browser access to LinkedIn. You may see permission prompts — please approve them. I'll wait."

Wait for pages to load successfully before continuing.

### 4. Spreadsheet Scripting (Basic plan only — skip for Pro)

**Google Sheets:** Open the Sheet and set up Apps Script so the browser permission popup is handled now (not mid-pipeline):

1. Navigate to the user's Google Sheet
2. Go to Extensions > Apps Script
3. Create a simple test function and run it:
   ```javascript
   function test() { Logger.log("GTM Kit connected"); }
   ```
4. **Tell the user in chat:**
   > "Google Sheets may show an authorization popup in your browser. Please approve it so I can write data to your Sheet during the pipeline run. Look for a popup or notification in your browser."
5. Wait for the script to execute successfully

**Excel Online:** Navigate to the spreadsheet URL and confirm you can edit cells.

**Excel Local:** Verify the file exists and is writable. Check that Python (`openpyxl`) or Node (`xlsx`/`exceljs`) is available via the Bash tool.

**Do NOT proceed until the spreadsheet test succeeds.**

### 5. Confirm Search Plan

Propose the first search you plan to run based on the ICP (target role, industry, geography) and **wait for user confirmation** before starting.

### 6. Confirm Ready

Tell the user:

> "Preflight complete — LinkedIn and storage are connected. I'll automate research and drafting, then pause for your approval before anything is sent."

---

## Step 1: Find Leads

Run the **find-leads** skill:
- Execute the confirmed search
- Extract and store new leads with `stage: "new"`
- Sync results per `references/storage.md`
- Report: "{n} new leads found"

## Step 2: Process New Leads

Run the **process-leads** skill:
- Fetch `stage: "new"` leads
- Qualify leads against ICP
- Move qualified leads to `stage: "qualified"` and weak-fit leads to `stage: "unqualified"` or `stage: "excluded"` with a 2-4 sentence `qualification_reason` based on ICP, product/pitch fit, evidence, and uncertainty
- Sync results per `references/storage.md`
- Report: "{n} leads processed"

## Step 3: Connect with Qualified Leads

Run the **connect-requests** skill:
- Send connection requests to all newly qualified leads
- Present personalized notes for explicit approval before sending
- Report: "{n} connection requests sent"

## Step 4: Message New Connections

Run the **message-new-connections** skill:
- Check recently accepted connections (since last run, up to 7 days back)
- Identify ICP matches, skip non-ICPs silently
- Send contextual DMs (different tone for outbound-accepted vs inbound-accepted)
- Sync results per `references/storage.md`
- Report: "{n} new connection DMs sent"

## Step 5: Message ICP Connections

Run the **message-connections** skill:
- Work through the "ICP Connections" tab queue (Status = "New")
- Present contextual DMs for explicit approval before sending
- Skip if queue is empty or if process-connections hasn't been run yet

From Kasper-AI/gtm-kit-plugin, MIT.

More from Kasper-AI/gtm-kit-plugin