patrol-briefinglisted

# Patrol Briefing A seat starts blank and inherits nothing. Every delegation failure is a briefing failure: the seat didn't know the constraint, the context, or what done looked like. A brief costs ~1–2k tokens; a wrong deliverable costs the whole seat-run plus the re-run. ## The six sections (all mandatory) Fill every one for the specific seat. If a section feels skippable, it is usually the one that would have prevented the failure. 1. **Context** — the 3–6 facts the seat can't discover cheaply: what project, what's already decided, what's already ruled out. What you'd tell a contractor on day one. Paste the two paragraphs that matter; don't point. 2. **Task** — one paragraph, imperative, singular. Two tasks = two seats. 3. **Why** — one sentence on what the output feeds. This lets the seat make the dozen micro-decisions you didn't anticipate correctly. 4. **Done looks like** — exact deliverable shape: format, length, fields, file paths to write, "return conclusions not raw tool output". Schema if you can give one. 5. **Do NOT** — the anti-scope: files not to touch, decisions not to make, rabbit holes not to enter. Seats over-help; fence them. 6. **Escalate if** — restated for this task: the specific ambiguity or above-tier trade-off that should come back rather than be guessed. ## Role slots Fill the bracketed slots per role; keep the six sections. - **executor** — model per routing (opus xhigh multi-step / sonnet med scoped / gpt-5.5 bulk). C