Roster Microsite Builder

Roster Microsite Builder

Spin up a good-looking, shareable website — a media kit for one creator or a whole roster — styled after the audience-report page, saved as a Cowork artifact for one-click sharing. Doubles as the agency "wrap deck". It is also a big generation engine: an agency rendering its roster pulls (and thereby seeds the pooled cache with) an audience report per creator in one move, so the next render next month is mostly free.

Read first: ${CLAUDE_PLUGIN_ROOT}/shared/conventions.md (tool schemas, credit prices, the nine conventions). This skill honors the credit-estimate discipline (${CLAUDE_PLUGIN_ROOT}/shared/credit-modes.md) and the Freshness Gate (${CLAUDE_PLUGIN_ROOT}/shared/freshness-gate.md). Use the creatorland-design skill for brand tokens, fonts, and components so the microsite is on-brand.

One creator, a procurement/legal dossier (not a website) → creator-diligence-audience-report. Just the cost estimate for a roster → run check_audience_coverage alone.

Inputs to collect

• Scope (required) — a single creator, or a roster (list / CSV / pasted handles). For each: platform + handle (and source_user_id if known — it guarantees a cache hit).
• View — single media-kit page, or a roster grid/index. Default: roster grid when >1 creator.
• Consent (creator-side use) — if this microsite is a MEMBER's OWN media kit (their data shown back to them or to brands on their behalf), confirm consent before exposing their data. Flag it; do not assume.

Flow

1. Scope + cost preflight (MANDATORY). Run check_audience_coverage { creators: [...] } (free) across the whole roster first, then state:

   creators: cached (free), fresh ≈ <F×25> credits (~$<F×0.63>). Build all, or only the cached ones, or cap to N? Continue?

   Pull only after the human confirms the spend. Honor "only the cached ones" and "cap to N".

2. Pull data per creator (cache-first) — get_audience_report { platform, source_user_id | handle } for each in scope (25 credits each, or $0 on a cache hit). Optionally query_market_intelligence { mode: "rate", vertical, creator_tier } per creator (derive creator_tier from that creator's follower count — emerging <1k / nano 1k-10k / micro 10k-100k / mid 100k-500k / macro 500k-1M / mega 1M+ — so the chip scopes to same-size creators; a tier too thin for the floor broadens, disclosed, to the vertical band) for a rate-band chip. Capture each report's as_of {profile_updated} and coverage block.

3. Render the microsite — ONE self-contained HTML artifact, Creatorland- branded (use creatorland-design tokens), styled after the audience-report layout:

   • Per-creator card/section: avatar (the paid-plan avatar { url, source } from search_creators/get_creator_profile — render the headshot when url is present, and degrade gracefully to an initials monogram when it's null, e.g. on free/demo plans), handle, platform, followers, real engagement rate, fake-follower % (suspicious/bot) + mass-follower % (real but low-value) + a red/green credibility badge, audience age/gender charts, geo (top countries / US weight), top interests
     • audience brand-affinity, recent/top posts, and brand history.
   • Roster view: a clean grid/index with filters (tier, platform, audience geo, credibility).
   • Charts: Chart.js loaded from CDN only — the artifact must be fully self-contained (no external storage APIs, no localStorage), per the Cowork artifact rules.

4. Save as a Cowork artifact (create_artifact) so it is shareable + re-openable. Offer a PDF export of the same page.

5. Honest data labels — per creator: "audience data as of "; a roster coverage line ("audience data available for of "); credibility badge labeled vendor-derived; Instagram-first coverage noted where TikTok/YouTube lack credibility / audience brand-affinity. Strip all PII — no contact info anywhere on the page.

Deliverable

A single shareable Cowork HTML artifact:

• Single creator: a one-page media kit (hero + the per-creator card above).
• Roster: a header (roster name, N creators, coverage "M of N", date) + a filterable grid of per-creator cards.
• Footer/provenance: "Audience data as of ; credibility scores are vendor-derived; no contact information shown (by design)."

Marketing tie-in: ship one live example generated by this skill, linked on the MCP marketing site as "see a sample roster site" (see CRE-889).

Honesty rules

• Cost is confirmed before pulling. The free check_audience_coverage preflight runs first; the paid pulls happen only on explicit go-ahead.
• Self-contained artifact. Chart.js from CDN only; NO external storage / localStorage / sessionStorage (Cowork artifact rules) — all data inline.
• Cache-first economics. A re-render next month is mostly cache hits (free vendor cost), so say " of were free from cache" in the build summary.
• Credibility is vendor-derived — badge it. Never imply the score is a Creatorland computation.
• Fake ≠ mass followers. Two figures, not one: fake-follower % = suspicious/ bot accounts; mass-follower % = real people who over-follow (low-value, not fake). Never merge them into a single "fake" number.
• Instagram-first honesty. TikTok/YouTube cards show demographics + geo and state credibility / audience brand-affinity are "not available for this platform today" — never fabricated.
• Consent gate for creator-side use. A member's own media kit is shown only after consent is confirmed.
• PII invariant (convention 7). No contact info on the page, regardless of source (the MCP strips the report's contacts; you add none). "Contact via Creatorland" is the only contact affordance.

Credit footprint

25 credits per creator needing a fresh report (cached creators are the same 25 billed but $0 vendor cost), + optional 5 per rate-band chip. A 12-creator roster with 5 cached ≈ 7×25 = 175 credits ($4.40); the same roster re-rendered next month, mostly cached, is far cheaper. The free check_audience_coverage preflight always runs first and the estimate is stated before any pull.