Zero Brief Discovery

Zero-Brief Discovery

No brief, no seed creator, just a brand. This skill bootstraps a slate from the brand's own footprint: it reads the creators already affiliated with the brand (pro affiliation data), picks the best-fitting ones as seeds, runs lookalikes off them, and proposes "more creators like the ones who already work with you." For brand-side managers staring at a blank page.

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).

If the user has a written brief, use brief-to-shortlist. If they name a single creator they love, use lookalike-ladder. This skill is for the blank-page case: the brand is the only input, and its existing affiliations are the bootstrap.

Precision & filters for outreach (convention 14). When this list will feed an outreach step, default to precision over raw recall: search precision: "tight" with the hard-gate filters the brief supports (platform, niche, data_freshness_days, content_format, audience_country), and reserve broad plus heavy lookalike unioning for market sizing. Do not fan out dozens of lookalike calls to hit a volume target (each hop drifts from the seed), and trim the weak tail by each result row relative_fit (within-set fit, 1.0 = strongest) rather than padding to a round number. Canon: ${CLAUDE_PLUGIN_ROOT}/shared/conventions.md convention 14.

Inputs to collect

• The brand (required) — name, and its vertical if not obvious.
• How they know existing partners (one path-choosing question) — the brand may already know a few creators who've worked with them. If they name some, those become the seeds directly (skip the affiliation-discovery step). If they don't, you discover affiliated creators via the corpus.
• Slate size — default 8.
• Plan check — affiliation data is pro-gated. If the workspace is free tier, affiliation discovery is unavailable; say so and offer the fallback (user names a seed → effectively lookalike-ladder, or provides a brief → brief-to-shortlist).
• Credit mode — default thorough; thrifty on trigger phrases.

Flow

Step 1 — Find the brand's affiliated creators. There is no brand-keyed search tool, so discover affiliations via the corpus: run a brief search scoped to the brand's vertical and use get_creator_profile's brand affiliations field (pro) to identify creators already tied to the brand. Practically:

search_creators {
  "mode": "brief",
  "brief": "<the brand's vertical and category, e.g. 'clean beauty / skincare creators'>",
  "filters": { "niche": "<brand vertical>", "country": "<home market if known>" },
  "limit": 25,
  "precision": "balanced"
}

Then profile candidates and keep those whose affiliations include the brand:

get_creator_profile { "identifier": { "type": "creatorland_user_id", "creatorland_user_id": "<id>" } }

If the user already named existing partners, skip the discovery search and profile those named creators directly. If NO affiliated creators surface (or free plan), disclose it plainly — "the corpus shows no creators affiliated with " — and offer to proceed from the vertical alone (labeled as a vertical slate, not an affiliation-grounded one) or to take a seed/brief from the user.

Step 2 — Pick the best seeds (no tool call). From the confirmed affiliated creators, choose the 1–3 best-fitting as lookalike seeds (strongest audience fit, freshest data, on-vertical). Show which seeds you picked and why.

Step 3 — Lookalike off each seed. One inference-free lookalike call per seed — exactly ONE of seed_creator / seed_content, never both, no filters:

search_creators {
  "mode": "lookalike",
  "seed_creator": { "type": "social_handle", "platform": "<platform>", "handle": "<seed handle>" },
  "limit": <2× slate size>,
  "precision": "balanced"
}

Multiple seeds = multiple lookalike calls; pool and dedupe the results. A creator surfacing across several seeds is a stronger pick — record the hit count as a ranking signal.

Step 4 — Profile the pooled candidates. get_creator_profile by returned identifier. Thorough: profile the deduped pool. Thrifty: profile the top slate size + 2.

Credit estimate fires here. Discovery search + affiliation profiling + lookalikes + candidate profiling typically exceeds ~30 credits — state the estimate and offer thrifty before fanning out (don't block).

Step 5 — Freshness Gate. Classify fresh / aging / stale; stale go to "re-verify before pitch" (thrifty: may drop, noting counts).

Step 6 — Conflict check (automatic here). Because the slate is built off the brand's own affiliations, flag any candidate ALSO affiliated with a competitor in the same vertical (from their profile affiliations), naming the competitor and the source. This is the whitespace-vs-conflict read the zero-brief approach makes natural.

Step 7 — One rate call for the vertical. Exactly one query_market_intelligence, wrapped in Refusal Recovery:

query_market_intelligence { "mode": "rate", "vertical": "<the brand's vertical>", "creator_tier": "<the candidate slate's tier, if it clusters at one>" }

Walk the ladder on refusal (thorough: until clearance; thrifty: max 2 rungs); disclose clearance level. When the discovered slate clusters at a creator tier, pass creator_tier (emerging <1k / nano 1k-10k / micro 10k-100k / mid 100k-500k / macro 500k-1M / mega 1M+) for a size-scoped band; when tier is mixed or unknown keep the vertical-wide band as the fallback and say so (a too-thin tier auto-broadens, disclosed, to the all-tier band).

Deliverable

# Discovery Slate — <brand>
_Built from who already works with you, <date> · <N> creators · Creatorland Data_

## Your existing partners, as the corpus sees them
<the affiliated creators used as seeds: @handle, why chosen — audience fit,
freshness. If none found: "No creators affiliated with <brand> appear in the
corpus; this slate is built from the <vertical> vertical instead — labeled
accordingly.">

## Discovery slate (ranked on resemblance to your partners)
### 1. <Name> — @<handle> (<platform>, <follower count>)
- **Why for you:** resembles <which of your partners> (embedding-based,
  inference-free) — <what they share: audience, content style, interests>
- **Seed hits:** <e.g. 2/3 of your seeds> · **Audience-geo:** <top geos>
- **Freshness:** fresh | aging (note)
- **Conflicts:** none | ⚠ also affiliated with <competitor> (per profile
  affiliations) — review before outreach
<repeat per creator>

## Indicative rate context (vertical band)
Market band for the **<vertical>** vertical<, broadened per note>:
p25 $<x> · median $<y> · p75 $<z> — <deal volume / recency>.
> <provenance line exactly as the tool returned it>
Market band for the vertical/tier — **not any individual creator's rate.**
<If broadened: disclose clearance level.>

## Re-verify before pitch
<stale creators; or "None.">

## Caveats
- The slate is grounded in your *known* affiliations in the corpus — partners
  not captured in Creatorland's data won't have seeded it.
- Resemblance is embedding-based (the data's read of your partners),
  inference-free — sanity-check the grids.
- Affiliation data is pro-gated; on free plan this skill can't run the
  affiliation bootstrap.
- Rate band is vertical-level; no per-creator rate data exists.
- No creator contact info is included or available via this tool; route
  outreach through Creatorland connections or the creators' public profiles.

---
Data freshness: <N>/<M> creators synced within the last sync window; <K> flagged for re-verification.
Provenance: Creatorland Data MCP · 1 discovery search + <S> seed lookalikes + <M> profiles + 1 rate benchmark (<clearance level>) · <date>.
Credits used this run: ~<N> (<1+S> searches ×2 + <M> profiles ×1 + <R> market-intel ×5).

Honesty rules

• The slate is grounded only in affiliations the corpus knows — disclose that partners outside the corpus didn't seed it; never imply complete coverage of the brand's real partner set.
• If no affiliations are found (or free plan), say so and relabel the output as a vertical slate — never present a vertical search as "based on who works with you."
• Resemblance is embedding-based and inference-free.
• Conflict flags name the competitor and cite the profile affiliation source.
• The rate band is vertical-level, never per-creator.
• Disclose Refusal Recovery broadening at the clearance level, always.
• Never imply access to creator contact info (convention 7).

Credit footprint

thorough: ~45 credits (1 discovery search ×2 + ~18 profiles ×1 scanned for affiliations + ~2 seed lookalikes ×2 + ~16 candidate profiles ×1 + 1 rate call ×5; +5 per refusal rung) · thrifty: ~24 credits (1 discovery search ×2 + ~10 profiles ×1 scanned + 1 seed lookalike ×2 + ~5 candidate profiles ×1 + 1 rate call ×5, max 2 rungs)