Brief and asset list (atom)

Load: ../plan-campaign/references/brief-template.md, ../plan-campaign/references/launch-and-calendar.md, the write-report skill (../write-report/SKILL.md) (§1 writing rules, §3 check).

Inputs: all earlier sections; approved claims = the claims that are usable through the claims-gate skill (../claims-gate/SKILL.md) ("Claims sign-off" marker, "Claims you may use" must facts, dated Brand record; or confirmed brain claims), with server applicability from knowledge_search ["claim"] + product_id; the angle bank and voc bank files (research/competitors/angle-bank.md, research/audience/voc-bank.md); existing campaign concepts; the brand voice from the kit.

Do:

1. Angles per Run/Test channel: one reason to buy each, resting on a named customer line and an exact approved claim (with id) whose applicability covers it (brand, or product for that product). unknown applicability is not brand-wide proof — the server refuses it on a concept; mark needs proof (scope unconfirmed). Adapt the expression per channel. No approved proof → needs proof, never invented. Press quotes and the brand's own site copy are not customer words; if research-market's voice-of-customer bank is empty, say so.
2. Asset list: one numbered row per asset — leg, placement, spec (format, ratio, length), angle, message/hook, claim (verbatim), CTA → destination (with UTMs), producer role (static ads, video ads, photos, content, human; video ads and photos are made with the goose-video and goose-product-photos skills, or by a person in local mode — the hand-off roles in ../plan-campaign/references/brief-template.md), due date (from the backwards calendar), dependency. Check the count against each paid leg's creative-volume need.
3. Measurement plan per leg: conversion event, UTM convention, promo code, attribution window, the store/CRM number it is judged on, blended MER target, "how did you hear" question.
4. Hand-offs: one row per leg, no silent legs; legs with no playbook are marked human-executed.
5. Replan triggers with thresholds and the action each starts.
6. Assemble the plan with the exact headings of the template, In short first; each section opens with its point in one sentence; plain words; no ids, codes, paths, tool or playbook names; every outside fact with its source and date. Fill every section or write "none — ". Then run write-report §3's check.

Output: the full brief_md text and the concept list (person, message, proof, evidence, confidence, product_id) for the write-back.

Never: a claim the claims gate does not allow; an asset with no owner or date.