GTM OS — Master Orchestrator

# GTM OS — The Master Index This is the orchestrator for a 6-file system that turns Claude into a GTM operator who knows your company. The other six skills in this family are: | Skill | File it owns | What it captures | |---|---|---| | `gtm-os-claude-md` | `CLAUDE.md` | Weekly index — current week, live campaigns, top priority | | `gtm-os-profile` | `profile.md` | Product, use cases, deal profiles, references, team | | `gtm-os-icp` | `icp-definition.md` | Buyer precision — fit signals, anti-ICP, qualification | | `gtm-os-positioning` | `positioning.md` | Value pillars, what-not-to-say, objection frameworks | | `gtm-os-competitor-radar` | `competitor-radar.md` | Win/loss patterns, competitive intel, monitoring | | `gtm-os-signal-library` | `signal-library.md` | Buying signals, hooks, performance log | These are designed to be used together. `CLAUDE.md` is the index every session reads first; the other five are the foundation it points to. The whole system lives in a single directory (typically `gtm-context/` or `.claude/gtm/`) inside the user's project. ## Your job as the orchestrator When this skill triggers, figure out which of three modes the user is in: ### Mode 1 — Setup from zero The user wants to build the system for the first time. Walk them through it in this order, one file at a time. Don't try to do all six at once — it's overwhelming and the answers degrade. Read each child skill when you get to it. 1. **profile.md first** — it's the foundation everything else references. Without it the others are abstract. 2. **icp-definition.md** — depends on profile (you need to know the product before you can define who buys it). 3. **positioning.md** — depends on both above. 4. **competitor-radar.md** — depends on positioning (you need to know your story before you can contrast it). 5. **signal-library.md** — depends on ICP (signals are useless without a buyer to apply them to). 6. **CLAUDE.md last** — it's the index, written after the things it points to exist. Tell the user upfront: "This is a 6-file system. We're going to build it one file at a time — typically over 30–90 minutes for the first pass, but you can spread it across multiple sessions. Each file will start as a draft you'll refine over weeks." Then ask which file they want to start with — default to profile.md if they have no preference. ### Mode 2 — Update an existing file The user references one specific file ("update my ICP", "add this competitor to the radar"). Don't load the orchestrator's full guidance — read the relevant child skill and let it drive. The orchestrator is just the routing layer. ### Mode 3 — Cross-cutting question or review The user asks something that touches multiple files: "review my whole GTM system", "is my ICP consistent with my positioning?", "what's the weakest file in my repo?". Read all six files (if they exist) and reason across them. Common cross-cuts: - **Profile ↔ ICP**: Does the deal profile by segment match the ICP definition? If profile says "ACV $80k average for mid-market" but ICP says "Series C+ only," there's a contradiction. - **ICP ↔ Positioning**: Do the value pillars actually map to the buyer's pain? Generic value pillars are the symptom of a vague ICP. - **Positioning ↔ Competitor radar**: Is the "what we don't say" list consistent with how you actually win against competitors? - **Signal library ↔ ICP**: Every signal should be tied to a specific ICP segment. Signals without a target buyer are noise. - **CLAUDE.md ↔ everything**: Does the weekly index reflect what's actually live in the other files? Stale indexes are worse than no index. ## Where the files live Default location: `gtm-context/` at the root of the user's project, or `.claude/gtm/` if they want it hidden. If the user hasn't decided, suggest `gtm-context/` — it's discoverable, it sits next to other context, and it doesn't pretend to be Claude-specific (the files are useful to humans on the team too). Inside that directory: ``` gtm-context/ ├── CLAUDE.md ├── profile.md ├── icp-definition.md ├── positioning.md ├── competitor-radar.md ├── signal-library.md └── archive/ ├── icp-2025-Q3.md └── positioning-launch.md ``` ## The maintenance rhythm - **CLAUDE.md** — every Monday morning, 5 minutes. Non-negotiable. - **competitor-radar.md** — after every closed deal (win or loss). One paragraph minimum. - **signal-library.md** — after every campaign concludes. Update the performance log. - **profile.md, icp-definition.md, positioning.md** — quarterly review, or whenever something material changes. ## Anti-patterns to flag 1. **"Series B SaaS companies" as an ICP.** That's a category, not a definition. 2. **A positioning file with no "what we don't say" section.** 3. **A signal library that's a list of guesses with no performance log.** 4. **A CLAUDE.md longer than 50 lines.** 5. **A repo that hasn't been touched in 30+ days.** ## Quick reference: which child skill for which question - "What does my company do?" → `gtm-os-profile` - "Who do we sell to?" → `gtm-os-icp` - "How do we talk about ourselves?" → `gtm-os-positioning` - "Who do we compete with?" → `gtm-os-competitor-radar` - "What signals matter?" → `gtm-os-signal-library` - "What's happening this week?" → `gtm-os-claude-md` - "Set up the whole thing" → stay here, walk them through in order