Campaign Package Builder

Campaign Package Builder

The travel-itinerary framing: one skill assembles a full client-ready proposal in four sections — destination (target audience/archetypes), accommodation (creator tiers with rate bands), excursions (the deal types the corpus shows are active), and price (a budget allocation). It composes the work of several narrower skills into one paste-into-deck artifact. It is the bundling skill; when the user wants only one section, route them to budget-allocator, brief-to-shortlist, or vertical-briefing.

Read first: ${CLAUDE_PLUGIN_ROOT}/shared/conventions.md (tool schemas, credit prices, the nine conventions — especially #9 multi-target slates). This skill honors thrifty/thorough credit modes (${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).

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

• Campaign goal + vertical (required) — the brief in a sentence; infer vertical from it.
• Target audience / archetypes (required) — one or more target profiles (e.g. "LATAM Gen-Z skincare" + "US millennial wellness"). A LIST of targets triggers convention 9: one search per target, per-target sections — never one merged search.
• Total budget (required) — drives the price section's allocation.
• Tier shape (optional, default macro/mid/micro mix biased by goal).
• Deal types (optional) — else inferred from the market-mode mix.
• Geo / platform constraints (optional).
• Mode — default thorough; thrifty on request.

Never ask twice; never ask for what the conversation already holds.

Flow

This skill composes budget-allocator + per-target discovery + market context. Convention 9 governs the audience section: N target archetypes = N searches and N per-target sections, plus an overview.

1. Market context (excursions + bands) — query_market_intelligence, each wrapped in Refusal Recovery:

   • { mode: "market", vertical: <vertical> } (floor 5/25) → which deal types are active (the "excursions") + deal-type mix.
   • { mode: "rate", vertical: <vertical>, deal_type: <dominant or requested>, creator_tier: <the accommodation (creator) tier being priced, when known — one call per tier> } (floor 10/50) → the rate band that prices accommodation tiers and feeds the budget split. When an accommodation tier's follower band is known, pass creator_tier (emerging <1k / nano 1k-10k / micro 10k-100k / mid 100k-500k / macro 500k-1M / mega 1M+) so each tier benchmarks against same-size creators; a tier too thin for the floor broadens (disclosed) to the vertical band. Disclose clearance level on any broaden. 5 credits each incl. retries. Thrifty: the rate call only, max 2 ladder rungs.

2. Per-target discovery (destination + candidates) — for EACH target archetype (convention 9), one search_creators: { mode: "brief", brief: <campaign goal + this target's description>, filters: { niche: <niche>, country: <geo>, platform: <if set>, min_followers: <tier floor>, max_followers: <tier ceiling> }, limit: <by tier need, cap 15>, precision: "balanced" } 2 credits each. Thin exact niche → thin-niche honesty ladder (convention 8): disclose, broaden one labeled step, tag picks exact/adjacent. Thrifty: smaller limit, fewer tiers per target.

3. Budget allocation (price) — no new tool calls; reuse the step-1 rate band to construct the tier mix exactly as budget-allocator does (tier planning rates = positions within the corpus band, labeled as such), fitted to the total budget, with the arithmetic shown.

4. Freshness Gate on named candidates — in thorough mode, get_creator_profile { identifier: { type: "creatorland_user_id", creatorland_user_id: <from search> } } (or social_handle) for the candidates that make the proposal; classify fresh/aging/stale; stale → "re-verify before pitch". 1 credit each. Thrifty: skip; label candidates "unverified — search-result freshness".

5. Assemble the package. Pre-run estimate fires before steps 2–4 if the plan exceeds ~30 credits (thorough multi-target runs do — say so).

Deliverable

# Campaign package — <campaign name> · <vertical>
## Overview
Brief: <one-line restatement>. Methodology: market context from the
Creatorland corpus; candidates from per-archetype searches; budget split from
corpus rate bands. Per-target caveats noted in each section.

## 1 · Destination (target audience)
<Per archetype — convention 9 sections; on paid plans front each candidate card
with the creator's `avatar { url, source }` headshot, initials fallback when null/free:>
### Archetype A — <description>
Candidates: @h1 (<fit>, <fresh/aging>, [exact/adjacent fit]) … 
### Archetype B — <description>
…

## 2 · Accommodation (creator tiers + bands)
| Tier | Follower range | Band (corpus, <deal type>) |
|---|---|---|
| Macro | … | $x–$z (med $y), ~upper band |
| Mid / Micro | … | … |
Bands are corpus-level market ranges for the vertical/tier — not any named
creator's rate.

## 3 · Excursions (active deal types)
<Deal-type mix from market mode, each a citation-ready block: stat +
provenance + recency window.>

## 4 · Price (budget allocation)
| Tier | Slots | Planning rate (band position) | Subtotal |
|---|---|---|---|
| … | | | |
| **Total** | | | **$≤budget** (<headroom>) |
Method note: planning rates are positions within the corpus band, not
follower-tier price lists.

## Re-verify before pitch
<stale candidates, or "none">

---
**Benchmark basis:** <clearance level(s) + floor-disclosure note where
broadened — "a privacy feature of the data source, not missing data">
**Provenance:** <provenance lines from each market-intel response, verbatim>
· recency window: <window>
**Data freshness:** N/M named candidates synced within the last sync window;
K flagged for re-verification. <thrifty: "candidates unprofiled — verify
before client use">
**Honesty note:** all rates are corpus-level market bands for the vertical/
tier — not personal rates. Per-target candidate lists disclose any niche
broadening. Confirm quotes before committing budget lines.
Credits used this run: ~N (breakdown: <market-intel>×5 + <searches>×2 + <profiles>×1)

Honesty rules

• Bands and budget rates are corpus-level for the vertical/tier — never a named creator's rate; the accommodation table and price method note both say so (convention 2).
• Multi-target = sectioned slate (convention 9): every archetype gets its own search and section; never one merged generic search across targets.
• Thin-niche honesty per target (convention 8): disclose, broaden one labeled step, tag exact vs adjacent fit.
• Disclose every broaden (Refusal Recovery clearance level).
• Budget arithmetic closes — subtotals ≤ budget, headroom shown; if the goal won't fit at band rates, say so.
• Freshness Gate is mandatory on named candidates.
• No contact info anywhere in the package, regardless of source (convention 7) — scrub bio snippets.

Credit footprint

thorough: ~30–55 credits (2 market-intel ×5 + N target searches ×2 + ~15–25 profiles ×1; refusal rungs add 5 each — pre-run estimate fires) · thrifty: ~12–18 credits (1 rate call + N smaller searches, no profiling, capped ladder)