---
name: os-install
description: Creates or repairs the shared workspace that every SCALE and GTM AI OS plugin depends on, detects an existing workspace and installed skills, and runs quick (3 questions) or advanced (12 questions) onboarding. Use when the user says "set up my GTM AI OS", "install the OS", "create the workspace", "I just installed the core plugin", "where do my files go", or when any module skill reports the workspace is missing.
---

# os-install

Build the workspace. Everything else in the OS reads from and writes to this folder.

## Step 0: Check what already exists
Before creating anything:
1. Look for an existing `GTM-AI-OS/` folder on the Desktop, in Documents, or at a path the user names. If found, read `INDEX.md` and `signals/log.md` and report: "You already have a workspace at <path> with <n> artifacts; last run was <skill> on <date>." Then ask: **Keep it** (continue with what's there), **Merge** (add any missing files and templates, never overwrite), or **Start fresh** (create a new folder next to it; never delete the old one).
2. Check which skills are already present. Read the last 50 lines of `signals/log.md` for skill names, and ask the user: "Which of these plugins do you have installed: core, scale-set, scale-chart, scale-align, scale-launch, scale-evolve, os-know, os-decide, os-act, os-learn?" Record the answer in `INDEX.md` under a `## Installed` heading so the router can route without guessing. If a skill the user names is not installed, tell them the plugin file to add.
3. If no workspace exists, continue to Step 1.

## Step 1: Ask where it lives
Ask one question: "Where should the GTM-AI-OS folder live? Default is your Desktop." Accept a path or "default".

## Step 2: Request access
If Cowork cannot see that location, tell the user exactly this: "Open the folder picker in Cowork and select the folder (or its parent) so I can create files there." Wait until access is confirmed. Never write elsewhere.

## Step 3: Create the structure
Create these files and folders. Do not overwrite an existing file; if one exists, read it and offer to merge.

```
GTM-AI-OS/
  CLAUDE.md
  INDEX.md
  context/company.md
  context/outcome.md
  context/icp.md
  context/deal-history.md
  context/signal-library.md
  context/messaging.md
  context/stack.md
  context/reversal-triggers.md
  context/pilot.md
  context/metrics.md
  context/governance.md
  context/fence.md
  artifacts/.keep
  artifacts/accounts/.keep
  signals/log.md
  evolve/.keep
  evals/README.md
```

Use the templates in `references/templates.md` for every file. Each context file starts with a status line `status: empty` that skills update to `status: draft` or `status: verified`.

## Step 4: Ask what they run
Ask: "Which systems hold your accounts and deals, your calls, and your email? Name whatever you use." Write the names to `context/stack.md`. Do not assume a product. For each system, tell the user which connector to enable in Cowork, or that pasting and uploads will work if no connector exists. Then ask which track they're starting: **SCALE** (internal AI transformation) or the **GTM AI OS** (external: who to go after). Both can run in one workspace.

## Step 5: Choose the onboarding depth
Ask: "Quick onboarding (3 questions, 5 minutes) or Advanced (12 questions, 20 minutes, better first results)?"

**Quick.** Ask one at a time:
1. "What does your company sell, to whom, and roughly how large is the revenue team?"
2. "What is your role and who do you report to?"
3. "If AI worked perfectly this year, which one number on the board deck would change?"
Write 1 and 2 to `context/company.md`, 3 to `context/outcome.md` as a draft outcome hint. Mark both `status: draft (quick)`.

**Advanced.** Ask one at a time; accept "skip" on any:
1. What you sell, to whom, and the revenue team size and structure (SDR, AE, AM, CS, marketing, RevOps headcount).
2. Your role, who you report to, and who your executive sponsor for AI is (or "nobody yet").
3. The one number you'd want to move, with today's value if you know it.
4. Your motions: inbound, outbound, partner, PLG, expansion. Which one matters most this year.
5. Your ICP in two sentences, and the segment you win most often.
6. Average deal size, sales cycle, and win rate, roughly.
7. Every system you run: CRM, calls, email, enrichment, sequencing, BI, chat. (Fills `stack.md` in detail.)
8. AI tools already in use and by whom. Any that were bought and dropped, and why.
9. Prior AI pilots: what was tried, what happened.
10. Constraints: regulated industry, security review requirements, data that can't leave the company.
11. Fiscal year and planning cadence, and any date the AI plan is due.
12. Your best guess at where work gets stuck most in the revenue process.
Write 1, 2, 4, 5, 6, 11 to `context/company.md`; 7 and 8 to `context/stack.md`; 3 to `context/outcome.md`; 9, 10, 12 to a new `context/history.md` (prior pilots, constraints, friction guess). Mark all `status: draft (advanced)`. Tell the user which SCALE or OS skill will verify each draft.

## Step 6: Log and hand off
Append to `signals/log.md`: `<date> | os-install | workspace created at <path> | company.md and stack.md drafted | next: scale-outcome-charter`.
Tell the user: SCALE track → "Install `scale-set` and say 'build my outcome charter'." OS track → "Install `os-know` and say 'run win loss'." Also: "The fence in `context/fence.md` is a draft until your revenue leader signs it. Nothing external sends before that."