Lookalike Ladder

Lookalike Ladder

Takes one beloved-but-unavailable seed creator and produces "same vibe, three price points": lookalike candidates profiled, sorted into follower-tier rungs, each rung framed against the vertical's market rate band. For casting directors whose client fell in love with talent the budget (or calendar) can't have.

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 AND a seed creator and wants the intersection, that's triangulated-casting. If they have only a brief, use brief-to-shortlist. This skill is seed-only: one creator in, a price-tiered ladder out.

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 seed (required) — a creator handle/URL, or a specific post URL if the client loved one piece of content rather than the creator overall. Exactly one seed; if the user names several loved creators, that's a multi-seed job — offer triangulated-casting per seed or run this skill once per seed as a slate (convention 9).
• Why the seed is out (optional, one question max) — "booked", "too expensive", or "just want options". Shapes the ladder emphasis (price-led vs availability-led framing); default to price-led.
• Ladder size — default 9 (3 per rung). Don't ask if context implies one.
• Geo / platform constraints (optional) — only if the user volunteers them; lookalike mode itself takes no filters, so these apply as a post-filter on profiled results, disclosed if they thin the ladder.
• Credit mode — default thorough; thrifty on trigger phrases.

Flow

Step 1 — Anchor profile. Profile the seed to learn its vertical, follower tier, audience geo, and interests — this anchors rung boundaries and the rate call:

get_creator_profile { "identifier": { "type": "social_handle", "platform": "<platform>", "handle": "<seed handle>" } }

(Use whichever identifier type the user gave — exactly one type per call.) Skip this step if the seed is a content URL with no known creator; infer vertical from the lookalike results instead, and say so.

Step 2 — Lookalike search. One inference-free call. Exactly ONE of seed_creator or seed_content — never both:

search_creators {
  "mode": "lookalike",
  "seed_creator": { "type": "social_handle", "platform": "<platform>", "handle": "<seed handle>" },
  "limit": 27,
  "precision": "balanced"
}

or, when the client loved a specific post:

search_creators {
  "mode": "lookalike",
  "seed_content": { "url": "<post URL>" },
  "limit": 27,
  "precision": "balanced"
}

Thorough: limit = 3× ladder size. Thrifty: limit = 2× ladder size, precision "tight" to cut noise early. Lookalike mode is embedding-based — no filters argument; any geo/platform constraints from intake are applied after profiling, with a note if they shrank a rung.

Step 3 — Profile fan-out. get_creator_profile per candidate, using the identifier type the search result returns:

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

Thorough: profile every candidate. Thrifty: profile the top ladder size + 3 by lookalike similarity.

Credit estimate fires here. Thorough at default size: 1 seed profile + 2 search + 27 profiles + 5 rate = ~35 credits — over the ~30 threshold, so state the estimate and offer thrifty before fanning out (don't block).

Step 4 — Freshness Gate. Classify every profiled candidate fresh / aging / stale. Stale candidates go only in "re-verify before pitch" (thrifty: may drop them, saying how many).

Step 5 — Build the rungs. Sort candidates into three follower-tier rungs relative to the seed: Premium (seed's tier or above), Mid (one tier down), Value (two tiers down / micro). Rungs are follower tiers, not prices — pricing context comes from step 6 and is band-level only.

Step 6 — One rate call for the ladder's vertical. Exactly one query_market_intelligence, wrapped in Refusal Recovery:

query_market_intelligence {
  "mode": "rate",
  "vertical": "<seed's vertical from step 1>",
  "creator_tier": "<the seed's follower tier — anchors the ladder; emerging <1k / nano 1k-10k / micro 10k-100k / mid 100k-500k / macro 500k-1M / mega 1M+>"
}

If refused (rate floor: 10 brands / 50 deals), walk the ladder per the module (thorough: until clearance; thrifty: max 2 rungs) and disclose the clearance level. When the seed's follower size is known, creator_tier scopes the band to the seed's tier as the ladder's anchor (a tier too thin for the floor broadens, disclosed, to the whole vertical); rung-level price differences are still follower-tier inference, never per-creator rates, and the deliverable says so.

Deliverable

# Lookalike Ladder — same vibe as @<seed>
_Seeded from <@handle | post URL>, <date> · <N> creators across 3 rungs · Creatorland Data_

## The seed, as the data sees it
<2–3 lines: vertical, tier, audience geo, the interest/vibe signals driving
the match — from the step-1 profile. If seed was a content URL: "matched on
the post's content embedding; vertical inferred from results.">

## Rung 1 — Premium (closest to @<seed>'s tier)
### <Name> — @<handle> (<platform>, <follower count>)
- **Vibe match:** <what the profile shares with the seed — interests,
  hashtags, audience shape; similarity is embedding-based, inference-free>
- **Audience-geo:** <top geos>
- **Freshness:** fresh | aging (note)
<repeat per creator; then Rung 2 — Mid, Rung 3 — Value, same row format>

## Price context (vertical band, not per-creator)
Market band for the **<vertical>** vertical<, broadened per note below>:
p25 $<x> · median $<y> · p75 $<z> — <deal volume / recency from the tool>.
> <provenance line exactly as the tool returned it>
Rung pricing logic: lower follower tiers typically transact lower in this
band — but the corpus has **no per-creator rates**; treat rungs as budget
directionality, not quotes. <If broadened: disclose clearance level per the
refusal-recovery module.>

## Re-verify before pitch
<stale candidates; or "None — all candidates cleared the freshness gate.">

## Caveats
- Similarity comes from stored content/profile embeddings — inference-free,
  but "vibe" is the model's read; sanity-check the grids before pitching.
- <any post-filters (geo/platform) applied and how many candidates they cut>
- No creator contact info is included or available via this tool; route
  outreach through Creatorland connections or the creator's public profiles.

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

Honesty rules

• Rungs are follower tiers; the band is vertical-level. Never write "Rung 3 costs $X" — write "value-tier creators typically transact lower within the vertical band". No per-creator rates exist.
• Lookalike similarity is embedding-based and inference-free — present it as "the data's vibe match", not a human editorial judgment.
• Disclose Refusal Recovery broadening at the clearance level, always.
• If post-filters (geo/platform) emptied a rung, say so — never quietly promote weaker matches up a rung (convention 8 spirit).
• Stale data never silently mixes into the rungs.
• Never imply access to creator contact info (convention 7).

Credit footprint

thorough: ~35 credits (1 seed profile + 1 search ×2 + ~27 profiles + 1 rate call ×5; +5 per refusal-ladder rung) · thrifty: ~22 credits (1 seed profile + 1 search ×2 + ~12 profiles + 1 rate call ×5, max 2 ladder rungs)