Chief of Staff Install

In your folder: brain/playbooks/decision-brief.md

Rendered from release 0.36.2, the version the download installs. This page is generated from the shipped file and says exactly what your copy says.

Playbook: Decision Brief

Run when the principal faces a decision and needs it framed for a call.

Frame it (don't stall on questions)

Settle these from memory — and where memory is silent, state your assumption and proceed. Only ask a question if the answer would flip the recommendation and you genuinely can't infer it; put the rest under Open questions in the brief rather than holding the brief back. (At interview level thorough — brain/role/principles.md → "The interview level" — ambiguity that would flip the recommendation may be asked first, one batch, before the brief; at every other level this playbook's deliver-in-this-turn contract holds.) Settle:

  • What exactly is being decided, and by when?
  • What's the objective / what does a good outcome protect or achieve?
  • What constraints apply (time, budget, people, reversibility)?

Produce

Produce the full brief in this turn — lead with the recommendation and deliver it directly, don't just offer to. If a decision is stated or implied, you also log it to memory/decisions.md with RAPID roles (per brain/playbooks/decisions.md) — but that logging is a side-effect that happens alongside the brief, never in place of it. A reply that only logs the decision and asks what to do next is a failure of this playbook.

Role coverage. If the decision clears the right-size bar, sweep for role holders as part of that logging step (brain/playbooks/decisions.md → "Role coverage"): run a specialist's mediated consult inline and fold the read into the brief, and leave the fuller work-up and any outreach draft as offers in this same turn. Neither offer holds the brief — the deliver-in-this-turn contract above governs.

  1. Decision — state it in one sentence, with the deadline.
  2. Recommendation — your recommended option, up front, in one line.
  3. Options — 2–4 realistic options. For each: what it is, key pros, key cons, and the main risk. Note which are reversible.
  4. Why the recommendation — the reasoning, tied to the objective and constraints.
  5. What it costs to be wrong — the downside of the recommended path and how to mitigate.
  6. Open questions — what's still unknown and how to resolve it, if that changes the call.

Principles

  • Make a real recommendation; don't hedge into a menu.
  • Flag irreversible options clearly — the bar for those is higher.
  • Check the deadline is real: urgency masquerading as importance is how a reversible decision steals an irreversible one's attention. Ask what waiting a week actually costs.
  • Keep the principal's stated risk tolerance and priorities in view.

After

  • Log the decision and rationale as a row in memory/decisions.md with RAPID roles once made — the same logging step as "Produce" above, confirmed done (per brain/schemas/capture-rules.md routing table; brain/playbooks/decisions.md is the single source of truth for the log's shape). A dated memory/log/ note may additionally record the event, but the decision + RAPID themselves live in memory/decisions.md so later briefs that read it actually surface it; update affected projects.