Playbooks — the single home for operational how-tos + tools
Why this exists: repeatable operational work (onboard a clinic, white-label an app, debug a data issue, run a sales motion) kept getting re-figured-out by each agent every time. This is the ONE place those live, so nobody re-derives steps and Atlas can hand a playbook straight to a Ready Practice staff member to work a client.
Each playbook is steps + the tools it uses + who does it (🤖 agent-only · 👤 staff-delegatable · 🧑💻 needs George/engineering). Four domains:
| Domain | For | Index |
|---|---|---|
| Onboarding | Get a new clinic (or white-label app) live end-to-end | onboarding.md |
| Customer Support | Respond to a client with Atlas, within SLA, escalate correctly | support.md |
| Debugging | Diagnose a production data/behavior issue safely | debugging.md |
| Sales | Run a lead → demo → close motion (Atlas-assisted) | sales.md |
The rule (so this never rots)
When you work out a repeatable operational process, write or update its playbook HERE — steps, the exact tool/script paths + callables, and the who-does-it tag. Do NOT leave it in an agent's head or a one-off issue comment. If a tool exists, link it; if it doesn't, note the gap. Atlas reads these to guide staff, so keep them staff-legible (plain language, no undefined jargon).
Where the pieces live (so a playbook can point at them)
- Reusable tools/scripts:
basis-functions/tools/(import/assign/ingest/backfill/audit/debug/verify),.github/scripts/(board, ship, lints),hybrid/sh/(builds, white-label flavor). - Deep technical references:
docs/claude/*.md(the CLAUDE.md topic index — firebase-backend, labs, scheduling, insurance-rcm, debugging-tools, etc.). - Canonical runbooks:
docs/claude/white-label-runbook.md,docs/handbook/engineering/*. - Test-as-staff pattern: callable-test-auth (custom token via appspot SA → idToken → call callables as staff) — see debugging.md.