Are playbooks just scripts?

They are script-like, but flexible. A playbook defines a default sequence of specialist calls and expected outputs. The CEO can skip steps, re-order them, or re-delegate based on what actually happens. It is a recipe, not a straitjacket.

Can I write a custom playbook?

Enterprise customers can contract custom playbooks today. Public authoring is on the roadmap. In the meantime, you can steer any open-ended task with a clear enough brief and the CEO will approximate the playbook you want.

Why does Landing Page Bootstrap take seven minutes?

Research (60s) + Content drafting (90s) + Code generation and deploy (180s) + Evaluator gate and fixes (60s) + browser screenshot and report (30s). The steps run partly in parallel; the headline number rounds up to a comfortable seven.

Can I pause a playbook?

Yes. The CEO will check in at milestones and will always respond to a "wait" or "stop" message. The circuit breaker also halts the playbook if any step trips a safety threshold.

What playbooks ship today?

Landing Page Bootstrap, Newsletter Send, Lead-List Enrichment, Inbox Triage, and Competitor Brief. More ship monthly. Scale and Enterprise plans get the full catalogue.

Docs · Core Concepts · Playbooks

Playbooks

When the CEO has done something a hundred times, it stops improvising. Playbooks are the pre-wired sequences that turn repeatable outcomes into reliable ones.

TL;DR

A playbook is a named sequence of specialist calls with a known good shape, a credit budget, and a success criterion. When the CEO recognises a matching goal, it runs the playbook instead of replanning from scratch. The owner sees a progress indicator, not a series of improvisations.

Why playbooks

Improvisation is expensive. Every novel plan the CEO writes burns tokens, risks a dead end, and costs the owner time. Most small-business jobs are not novel — they are "build a landing page," "send the newsletter," "triage the inbox," "enrich this lead list." Playbooks encode the CEO's accumulated experience into a shape that just runs.

Playbooks also make pricing predictable. The Landing Page Bootstrap playbook costs about 45 credits; we can tell you that because it has run thousands of times and we know. A custom plan could cost twenty or eighty; we would not know until it was done.

Landing Page Bootstrap, end to end

The canonical example. A new owner finishes onboarding, says "ship my landing page," and the CEO runs the bootstrap playbook. The steps:

Research Specialist (~60s). Pulls the owner's niche context from memory, does a quick competitor scan, drafts a messaging brief.
Content Specialist (~90s). Takes the brief, produces hero copy, two value-prop blocks, a call-to-action, SEO title + description.
Coding Specialist (~180s). Scaffolds an Astro site, wires the copy in, generates a GitHub repo, deploys to Railway or Vercel. Returns the live URL.
Evaluator gate (~60s). Checks the deploy, runs an accessibility lint, confirms the page renders. Sends it back if anything looks wrong.
Browser Specialist (~30s). Takes a screenshot of the live page, attaches it to the report card.

Total wall-clock: seven minutes including parallel overlap. Total credits: about 45. The owner sees one progress card ("Landing Page: step 3 of 5") and one final report card ("your landing page is live at [URL]").

Anatomy of a playbook

A trigger. The goal pattern that matches — "ship my landing page," "launch the newsletter," etc. The CEO's planner checks incoming goals against playbook triggers first.
A default sequence. The list of specialist calls, in order, with expected input/output shapes between them.
Credit and time budgets. What the playbook typically costs. Used both for the owner-facing estimate and as a circuit-breaker hint.
Success criterion. What "done" looks like — a live URL, a sent send, an archived inbox — so the Evaluator can check.
Skill pack hooks. Which skill packs the playbook prefers when available.

When playbooks bend

A playbook is not rigid. If the Research Specialist comes back with "the owner has not actually described their niche," the CEO pauses the playbook and surfaces a decision_required card to the owner. If the Coding Specialist cannot deploy because Railway credentials expired, the CEO re-plans to prompt the owner for reauth instead of blindly retrying. The sequence is a default, not a straitjacket.

Playbooks that ship today

Landing Page Bootstrap. Research → Content → Coding → Evaluator → Browser. Seven minutes to a live URL.
Newsletter Send. Content → Evaluator → Business Ops (through Resend). About four minutes.
Lead-List Enrichment. Research (parallel per lead) → Content → Business Ops. Scales with list size.
Inbox Triage. Business Ops reads the inbox; Content drafts replies; Evaluator flags anything that needs human review. Ten minutes a day, consistently.
Competitor Brief. Research (parallel per competitor) → Content. A readable brief instead of ten open tabs.

More playbooks ship monthly. Scale and Enterprise plans get the full catalogue on day one.

How it works in Black Box

Playbooks are registered through the CEO's playbook tool in apps/engine/src/ceo/playbook-tool.ts. Each playbook is a JSON-ish definition: trigger, step list with the delegation tool to call, expected duration, credit budget, and success criterion. At plan time, the CEO asks the playbook tool "is there a playbook for this goal?" and, if yes, runs the registered step sequence via the usual delegate_to_* tools.

The UI renders a playbook progress card — the "step 3 of 5" strip you see in the Boardroom — driven by the plan_proposed / plan_execution_started / plan_execution_finished events (see structured cards).

Frequently asked

Are playbooks just scripts?: Script-like but flexible. The CEO can skip, re-order, or re-delegate.
Custom playbooks?: Enterprise today; public authoring on the roadmap.
Why seven minutes?: Research + content + code/deploy + Evaluator + screenshot, with some parallel overlap.
Can I pause?: Yes — say "wait." Circuit breaker halts on safety trips.

Related concepts

Delegation — the primitive playbooks are built on.
Skill packs — what playbooks pull in for vertical work.
Evaluator gate — every playbook step is reviewed.
Feature page: Landing Page Bootstrap.
Glossary: agent playbook.

Next: the four views

Playbooks run in the background. Here is the UI surface where you watch.

The four views All concepts