Brief Builder

Brief Builder

Takes a call transcript (client call, internal kickoff, voice-memo dump) and produces the artifact agencies circulate for sign-off: a formal creator brief document — polished markdown a casting lead can send to the client or the internal team as-is. The output is deliberately shaped to feed brief-to-shortlist afterward, but the brief itself is the deliverable here.

Read first: ${CLAUDE_PLUGIN_ROOT}/shared/conventions.md (tool schemas, credit prices, the seven conventions). This skill honors thrifty/thorough credit modes and Refusal Recovery (relevant only to the optional enrichment call).

Calibration: if ~/.claude/plugins/config/creatorland/creatorland-data/PROFILE.md exists (written by /intake), read it first — persona, familiarity (1–3), and saved defaults set narration level and assumed context (conventions.md §15).

Trigger discrimination: the user wants a brief document out of a transcript. If they want creators found from a transcript, use transcript-to-shortlist. If they already have a written brief and want a shortlist, use brief-to-shortlist.

Inputs to collect

• The transcript (required).
• Whether to enrich the budget section with market norms — only offer this if a vertical is identifiable and the user is on pro; phrase it with its cost: "I can ground the budget section in market norms for with one market-intelligence call (5 credits) — want that, or keep this run at 0 credits?"
• Anything the gap questions surface — one batched round, max 4.

Flow

Step 1 — Transcript Intake (0 credits). Run ${CLAUDE_PLUGIN_ROOT}/shared/transcript-intake.md in full: extract the schema (brand/client, objective, audience, creator_spec, vertical, deliverables, compensation_type, budget, timeline, exclusions), tag every value stated / inferred / missing, then ask gap questions in ONE batched message, max 4, ordered by impact. For a brief document the high-impact gaps skew toward deliverables count/format, compensation type, timeline, and budget posture (the search-side four matter less here). Never ask what the call answered — quote it back. Capture named creators ("someone like @x") as reference talent for the creator-spec section, and competitor mentions for the exclusions section.

Step 2 — Confirm the requirements block. Show the normalized block with stated/inferred labels; incorporate corrections.

Step 3 (optional, pro, +5 credits) — Market-norms enrichment. Only if the user accepted in inputs. One call:

query_market_intelligence {
  "mode": "market",
  "vertical": "<the brief's vertical>",
  "deal_type": "<the brief's compensation type mapped to a corpus deal type, e.g. \"affiliate\" — omit if unmapped>"
}

Wrapped in Refusal Recovery (market floor: 5 brands / 25 deals; thrifty: max 2 ladder rungs). Use the result ONLY to add a disclosed "market context" note under the compensation section — counts/distributions for the vertical, with the tool's provenance line and the clearance level if broadened. Never let it overwrite what the client actually said about budget.

Step 4 — Render the brief document (template below). Stated values are written as client requirements; inferred values are written normally but flagged in the Open Questions section; missing-and-unanswered values become explicit placeholders, never invented content.

Step 5 — Offer the next step. End with: "Want me to run this brief through casting? Say 'build the shortlist' and I'll hand it to brief-to-shortlist as-is." (That run has its own credit footprint — see that skill.)

Deliverable

# Creator Brief — <brand/client> · <campaign working name>
_Drafted from the <date> call · for internal sign-off · v1_

## 1. Background & Objectives
<what the campaign is for and what success looks like — awareness / launch /
conversion / UGC volume / event, in the client's own framing>

## 2. Target Audience
<demographics, geo markets, language, community — bullet per dimension>

## 3. Creator Spec
- **Platforms:** <…>
- **Follower tier(s):** <…>
- **Vibe / content style:** <the call's own words, quoted where strong>
- **Reference talent:** <creators named on the call, e.g. "in the vein of @x"> 
- **Vertical:** <corpus vertical>

## 4. Content Deliverables
<type × count table, e.g. 3 IG reels + 1 dedicated YouTube video + 4 stories;
usage/whitelisting terms if discussed, else a placeholder>

## 5. Compensation
- **Comp type:** <job-suite tag: gifting+paid | affiliate+paid | pure affiliate | flat fee>
- **Budget:** <total and/or per-creator band, or "to be discussed">
<If enriched:>
> Market context (<vertical><, clearance level if broadened>): <counts /
> distributions as returned>. <provenance line exactly as returned.>
> Corpus-level context for the vertical — not a quote or a rate for any
> specific creator.

## 6. Timeline
<launch dates, flight windows, content deadlines — placeholder rows for
anything undiscussed>

## 7. Exclusions & Conflicts
<competitor conflicts named on the call, content restrictions, geo
exclusions; "none raised on the call" if so>

## 8. Approval Workflow
_Placeholder — to be completed at sign-off:_
- Creator approval: <name/role> · Content approval rounds: <n> · Final
  sign-off: <name/role>

## Open Questions
<every inferred value restated as "we inferred X — confirm?", plus any
missing fields the user chose not to answer>

---
Source: call transcript provided <date>; values marked inferred were derived,
not stated — see Open Questions.
<If enriched:> Market context: Creatorland Data MCP, 1 market-intelligence
call (<clearance level>) · <date>.
Credits used this run: 0 <or "~5 (+5 per refusal-ladder rung): 1 market-intel call">.

Honesty rules

• Never invent client requirements. Missing = placeholder; inferred = flagged in Open Questions. The sign-off artifact must be auditable against the call.
• Market enrichment is vertical-level corpus context, never a rate for any creator and never a substitute for the client's stated budget; always disclosed with provenance and clearance level.
• Budget "to be discussed" is a valid, honest value — don't pressure a number into existence.
• No creator contact info exists in this surface; reference talent entries are names/handles only (PII discipline, convention 7).

Credit footprint

zero-tool-call mode (default): 0 credits · with market enrichment: ~5 credits (1 market-intel call; +5 per refusal-ladder rung, thrifty max 2 rungs). The downstream shortlist run, if requested, is billed under brief-to-shortlist's footprint.