Creator Brand Matchmaker

Creator-to-Brand Matchmaker

The talent-manager side of the corpus. Given a creator, this skill builds a lookalike cohort (creators like them), reads which brand verticals and categories show the most deal activity around that cohort shape, and writes a pitch-target memo: which categories to aim outbound at, ranked by corpus deal activity, with provenance. For talent managers and creator-side reps prepping outbound, and for Creatorland's own creator-side story.

Read first: ${CLAUDE_PLUGIN_ROOT}/shared/conventions.md (tool schemas, credit prices, the nine conventions). This skill honors thrifty/thorough credit modes (${CLAUDE_PLUGIN_ROOT}/shared/credit-modes.md), Refusal Recovery (${CLAUDE_PLUGIN_ROOT}/shared/refusal-recovery.md), and the Freshness Gate (${CLAUDE_PLUGIN_ROOT}/shared/freshness-gate.md).

The honesty spine of this skill (read before building): the corpus shows DEAL ACTIVITY PATTERNS — which verticals/categories are transacting with creators of this shape — not brand intent. "Beauty is an active category for creators like yours" is true and useful; "Brand X wants to work with you" is NOT something this data can say. The memo targets CATEGORIES and VERTICALS, never specific brands as warm leads, and never names a brand contact or any contact info (convention 7). It's a "where to aim" memo, not a lead list.

If the user wants creators FOR a brand (the buy side), use brief-to-shortlist. If they want to benchmark a creator's quote, use fair-price-brief / one-number-rate.

Inputs to collect

• The creator (required) — an identifier (social_handle, creatorland_user_id, source_user, email, or phone). This is the manager's own talent or a creator they represent.
• Markets / geo (optional) — to weight categories by where the creator's audience is.
• Any categories to include/exclude (optional) — e.g. "we don't do alcohol or gambling."
• Credit mode — default thorough; thrifty on the usual trigger phrases.

Flow

Step 1 — Profile the seed creator. get_creator_profile { "identifier": { "type": "<chosen>", ... } } → the creator's vertical signals (interests, hashtags), audience geo, follower tier, freshness, and existing brand affiliations (pro). Run the Freshness Gate on this profile immediately: if the seed itself is stale, flag that the whole memo rests on aging data and recommend a re-sync before pitching off it.

Step 2 — Build the lookalike cohort. Seed a lookalike search off the creator to define "creators like them":

search_creators {
  "mode": "lookalike",
  "seed_creator": { "type": "creatorland_user_id", "creatorland_user_id": "<seed>" },
  "limit": 20,
  "precision": "balanced"
}

(thorough: limit 20; thrifty: limit 10.) This cohort is the bridge from one creator to a corpus-sized population whose deal patterns market-intel can read.

Step 3 — Determine the cohort's vertical(s). From the seed profile + cohort signals, identify the 1–3 corpus verticals the creator plausibly sits in (Beauty, Fashion, Health & Fitness, Food & Beverage, Technology, …). These drive the market-intel calls; show them back so the manager can correct.

Step 4 — Market-intel scan per candidate vertical (the core). One query_market_intelligence market-mode call per candidate vertical to read deal-activity shape — which company types / sub-categories are most active:

query_market_intelligence {
  "mode": "market",
  "vertical": "<candidate vertical>",
  "active_since": "<ISO date, recent window if weighting toward current activity>"
}

Each is wrapped in Refusal Recovery (market floor: 5 brands / 25 deals): if a vertical slice floors, walk the ladder (thorough: to clearance; thrifty: max 2 rungs) and disclose at what level it cleared. Read the returned distributions to rank brand categories/company-types/sub-categories by deal activity for this creator shape.

Credit estimate fires here. Thorough is ~1 profile + 1 lookalike search + ~10–20 cohort profiles (optional, see thrifty note) + 2–3 market-intel calls. If cohort profiling is on, the run exceeds ~30 credits — state the estimate, offer thrifty (which skips cohort profiling and reads categories from the search + market-intel layer only), proceed.

Step 5 — Cross-check against the creator's existing affiliations. From the seed profile's affiliations (pro), note categories the creator is ALREADY active in (exclude or de-prioritize as "already covered") vs adjacent categories that are active for the cohort but not yet for this creator (the prime whitespace pitch targets). Honest label: "already affiliated" vs "active for similar creators, not yet for you."

Step 6 — Write the pitch-target memo.

Deliverable

# Pitch-Target Memo — <creator name / @handle>
_Prepared <date> · Creatorland Data · deal-activity patterns, NOT brand intent (see how to read this)_

> **How to read this memo:** these are brand CATEGORIES and VERTICALS where
> the corpus shows active deal-making with creators like <name> — ranked by
> deal activity, not by any brand's stated interest. This tells you where to
> AIM outbound, not who is waiting to hear from you. No specific brands are
> named as leads, and no contact information exists in this data or this memo.

## The creator, as the corpus sees them
- **Shape:** <vertical(s), audience geo, follower tier from the profile>
- **Already affiliated in:** <categories from profile affiliations, or "none in corpus">
- **Freshness:** fresh | aging (note) — <if stale: "seed data is stale; re-verify before pitching off this">

## Pitch-target categories (ranked by corpus deal activity)

### 1. <Category / vertical / company-type> — high activity
- **Why it fits:** <tie to creator shape + cohort signal>
- **Activity basis:** <deal volume / company count for this slice + the provenance line exactly as the tool returned it>
- **For you specifically:** active for similar creators · <"and you're already in it" | "whitespace — not yet in your affiliations">
- **Benchmark basis:** <vertical-level / broadened level per Refusal Recovery, if applicable>

<repeat per ranked category, 3–6 entries>

## Whitespace (active for your cohort, not yet for you)
<the adjacent-category shortlist — where outbound is most likely to be net-new>

## Already covered
<categories the creator is already affiliated with — de-prioritized for outbound>

## How to use this
- Aim outbound pitches at the top categories; build the brand list yourself
  from your own network/research — this memo deliberately names none.
- Pair with a fair-price benchmark (`fair-price-brief`) before you quote.

## Caveats
- **Deal-activity patterns, not brand intent.** A hot category is not a
  warm lead. No brand here has expressed interest in this creator.
- Rankings are corpus-level for the vertical/category — not this creator's
  win probability.
- Refusal Recovery broadening disclosed at the clearance level above.
- No brand contacts, no contact information — by design (convention 7).
- <seed staleness / thin cohort caveats, if applicable>

---
Data freshness: seed profile <fresh|aging|stale>; <C>/<D> cohort creators synced within the last sync window (if cohort profiled).
Provenance: Creatorland Data MCP · 1 seed profile + 1 lookalike search<+ <C> cohort profiles> + <V> market-intel calls (<clearance levels>) · <date>.
Credits used this run: ~<N> (1 profile + 1 search ×2 <+ <C> profiles> + <V> market-intel ×5).

Honesty rules

• Deal-activity patterns, not brand intent — the load-bearing rule. Never imply a brand wants this creator; never convert a hot category into a warm lead.
• Never name specific brand contacts. No contact info, ever (convention
  7. — the memo targets categories/verticals only and says so explicitly.
• Rankings are corpus-level for the category/vertical — never "this creator's odds" or "their rate."
• Disclose Refusal Recovery broadening at the clearance level on every market-intel call.
• Distinguish "already affiliated" from "whitespace" honestly; don't sell a category the creator already works in as a fresh opportunity.
• If the seed profile is stale, the memo says its whole basis is aging — the Freshness Gate applies to the seed, not just a list.
• Never strip the provenance line to make the memo prettier.

Credit footprint

thorough: ~30–45 credits (1 seed profile + 1 lookalike search ×2 + ~10–20 cohort profiles ×1 + 2–3 market-intel ×5; +5 per refusal-ladder rung) — the estimate fires when cohort profiling is on · thrifty: ~13–18 credits (1 seed profile + 1 lookalike search ×2, cohort profiling skipped, 2 market-intel ×5, max 2 ladder rungs each).