crm-setup

---

name: crm-setup description: Stand up HQ's native CRM for a company, end to end — connect the services it already uses or build native HQ capabilities for the ones it doesn't, then turn the CRM on. Use when the user says "set up the CRM", "onboard to the CRM", "turn on CRM for ", "get our pipeline into HQ", or is starting client management in HQ. Service-agnostic: no vendor is required. Writes companies/{co}/client-management.yaml, wires each GTM silo via /new-connection or a native-entity capability, enables crmEnabled, syncs, and verifies /crm returns data. allowed-tools: Bash, Read, Write, Edit, Grep, AskUserQuestion, Skill, Agent

/crm-setup — stand up the CRM for a company

Walk a company from zero to a working canonical CRM. The CRM is HQ's ontology used as the canonical client record; external systems are optional enrichment. So this works whether the company has a full GTM stack or no tools at all.

The promise: if you use a tool, we connect it; if you don't, we build the capability natively in HQ. No vendor is required.

Prerequisites

• A resolved company slug {co} (ask if not given).
• The caller is an owner/member of the company (CRM enable is owner-gated).
• The hq CLI on PATH (for activation + sync).

Step 1 — Frame the silos

The CRM joins up to six GTM silos onto each client record. Explain them plainly and find out, for each, whether the company uses a tool:

Silo	What it holds	Common tools
inbound	where leads originate	a lead DB, a demo-request form, a survey DB
pipeline	CRM of record: stage, owner	Attio, HubSpot, Salesforce, a spreadsheet
contracts	agreements / signing state	PandaDoc, DocuSign
billing	invoices / subscriptions	Stripe, QuickBooks
mail	outreach + replies	Gmail, Outlook
meetings	call notes → activity timeline	HQ meeting notes, Recall, Fireflies

Use AskUserQuestion (one or two grouped questions) to learn, per silo: tool, or none? Don't force a tool — "none" is a first-class answer and routes to the build-native path.

Step 2 — Per silo: connect or build

For each silo the company cares about:

• Has a tool → connect. Invoke /new-connection (core) scoped to the company for that system. It picks the fastest harness, stores creds safely, writes companies/{co}/connections/{slug}.yaml, and returns a connection slug. Record the silo → connection slug mapping.
• No tool → build native. The company will maintain that silo's records directly in the ontology — no vendor. Scaffold a thin native-entity capability (a connect-{silo} skill or a /newworker worker) that creates/updates the relevant entities via the write path:

  hq crm entity upsert --company {co} --type <company|deal|contract|invoice> \
    --name "<name>" [--json <file>]
  

  hq crm entity upsert writes through the CRM write-gate and refreshes the projection, so native records are immediately queryable by /crm. This is the zero-vendor path that makes the CRM usable with no external services.

Default every silo to read-only / least-privilege; writes are confirmation-gated (the /new-connection and secure-sidecar discipline).

Step 3 — Write the config

Write companies/{co}/client-management.yaml capturing: the company, the owner identity (for "my action items / next steps"), each silo's mode (connected|native) + connection slug or native source, the pipeline stages vocabulary, and the secret key names in play (never values). Schema + field notes: the client-management-config knowledge doc.

This file is the parameterization seam every bundled client-management skill reads — it is what makes them company-agnostic.

Step 4 — Ensure there are accounts to project

The projection needs at least company account entities to show anything.

• If a pipeline/inbound silo is connected, the company's adapter (or the connection capability) supplies accounts.
• If silos are native, create a couple of company entities now via hq crm entity upsert so the projection isn't empty.

Step 5 — Turn the CRM on

CRM is off by default, gated per company by crmEnabled (owner-callable). Enable it from the terminal:

hq company settings set --company {co} --crm-enabled true

(Equivalent to the owner flipping it in the HQ console. Requires owner role.)

Step 6 — Sync and verify

1. /hq-sync for the company so the projection (crm-projection.json) builds + lands in the vault.
2. Run /crm — confirm it returns accounts (by stage) and a real client's joined record. If the projection is missing, re-check: accounts exist? crmEnabled on? sync ran? Report which, rather than hand-joining sources as a silent fallback.

Step 7 — Report

Summarize plainly: which silos are connected (and to what) vs native, that CRM is enabled, and that /crm is live. Point the user at the bundled skills they can now use (signals, action-items, meeting-next-steps, deal-brain, client-engagement) — all driven by the config you just wrote.

Rules

1. No vendor is required. "None" is a valid answer for any silo and routes to the build-native path. Never block setup on a tool the company doesn't have.
2. Connect via /new-connection, don't hand-roll. Reuse the core connection command for every connected silo so creds, capability, and the connection definition are handled consistently.
3. Native writes go through hq crm entity upsert. Don't write entity markdown by hand and hope the projection updates — it only refreshes on the write path.
4. Least-privilege, read-first. Read access by default; gate writes.
5. Secrets only via hq. Never paste or print a secret; /new-connection owns the safe credential flow.
6. Per-company isolation. Everything written stays in {co}'s scope; never pull another company's CRM data into setup.
7. Verify before declaring done. Setup isn't complete until /crm returns real data for the company.

Reuse map

Step	Use
Connect a silo's service	/new-connection (core)
Build a native silo	hq crm entity upsert + /newworker (capability)
Enable CRM	hq company settings set --crm-enabled true
Sync + verify	/hq-sync, then /crm
Config schema	client-management-config knowledge doc