Creator Diligence Report

Creator Diligence Report

The "credit bureau" artifact: a standardized one-pager per creator that procurement and legal teams will actually accept — every claim sourced, every gap labeled, nothing vibes-based. For agency ops leads and brand-side teams clearing a creator before contract. The deliverable is a fixed-format markdown one-pager (the same sections every time — standardization IS the feature).

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 live quoted fee to evaluate, that's fair-price-brief. If they only want competitor-relationship screening, that's conflict-check. This skill is the full standardized dossier.

Inputs to collect

• Creator identifier (required) — handle, profile URL, Creatorland ID, or email. Map to exactly one of the five identifier types (social_handle, creatorland_user_id, source_user, email, phone). Never ask for a format the user already gave in another form.
• Engagement context (optional) — the brand/campaign this diligence is for. Powers the "fit" framing and lets conflict-check-style screening run against named competitors if the user supplies them.
• Competitor brands to screen (optional) — ask once; skip if already named or the user says it doesn't matter.
• Credit mode — default thorough; thrifty on the standard triggers.

Flow

1. Profile — get_creator_profile { "identifier": { "type": "social_handle", "platform": "<platform>", "handle": "<handle>" } } (or the matching single identifier type). Extract: interests, hashtags, audience geo, follower count/tier, data freshness (Freshness Gate classification goes in the report header — a stale dossier must say so at the top, not the bottom), and brand affiliations (pro plan; if absent, the affiliations section states "unavailable on this plan" — never blank). 1 credit.

2. Benchmark position — query_market_intelligence { "mode": "rate", "vertical": "<creator's inferred primary vertical>", "creator_tier": "<the creator's follower tier, when known>" } — wrapped in Refusal Recovery (rate floor 10 brands / 50 deals; thorough: walk the ladder to clearance; thrifty: max 2 rungs). When the creator's follower size is known, pass creator_tier (emerging <1k / nano 1k-10k / micro 10k-100k / mid 100k-500k / macro 500k-1M / mega 1M+) so the band scopes to same-size creators — a truer read of where the creator's tier sits; a tier too thin for the privacy floor broadens (disclosed) to the all-tier band. This positions the creator's TIER within the vertical's market band — it is corpus context, never the creator's rate history. 5 credits per call incl. ladder retries.

3. Lookalike context (thorough only) — search_creators { "mode": "lookalike", "seed_creator": { "type": "social_handle", "platform": "<platform>", "handle": "<handle>" }, "limit": 5 } — names the creator's in-corpus comp set ("comparable talent" section). Inference-free, 2 credits. Do NOT profile the comp set. Skip in thrifty.

4. Compose the one-pager (no tool calls). Fixed section order, every section present even when empty (labeled "no data in corpus" / "supply from your own tools"). Engagement metrics, per-post performance, and sentiment have NO corpus source — those sections are labeled placeholder slots for the customer's own analytics, per the diligence concept's honest constraint.

Credit estimate: never exceeds 30 credits for a single creator, so the pre-run estimate doesn't fire; multi-creator diligence requests should route through roster-enricher or be run as repeated single reports with a stated N×13 estimate.

Deliverable

# Creator Diligence Report — @<handle>
_Creatorland Data · <date> · standardized format v1_
**Data freshness status: <FRESH | AGING | STALE — re-verify before contracting>**

## 1. Identity & platform
Handle / platform / follower count & tier / verification status (as returned).

## 2. Content profile
Interests: <from profile> · Hashtags: <from profile> · Inferred primary
vertical: <vertical> (inferred from interests — labeled inference).

## 3. Audience geography
<top geos with shares, exactly as the profile returned them>

## 4. Brand affiliations (corpus-derived)
<list, or "unavailable on this plan", or "none found in corpus">
Note: affiliations are derived from the Creatorland corpus — they are not an
exhaustive deal history. Absence here ≠ absence of relationships.
<If competitor screen requested: per-brand verdict — "no overlap found IN OUR
CORPUS" / "⚠ corpus shows affiliation with <brand>">

## 5. Market position (vertical band context)
Market band for the **<vertical>** vertical<scope note if broadened>:
p25 $<x> · median $<y> · p75 $<z>. The creator's <tier> tier sits <framing>.
> <provenance line verbatim from the tool> · recency window: <window>
This is the corpus-level band for the vertical/tier — NOT this creator's
rate history (no per-creator rate data exists).

## 6. Comparable talent (in-corpus)        [thorough mode]
@<c1>, @<c2>, @<c3>, @<c4>, @<c5> — lookalikes by stored embeddings.

## 7. Performance metrics                  [placeholder — no corpus source]
Engagement rate / per-post metrics / growth: ❏ supply from your own
analytics tools. The Creatorland Data MCP has no per-post metrics endpoint;
this report leaves labeled slots rather than inventing numbers.

## 8. Contact
Creatorland Member: <yes/no if known> — contact via Creatorland (connections
coming soon) or the creator's public profiles. No contact information is
included in this report (by design).

---
**Provenance:** Creatorland Data MCP · 1 profile + 1 rate benchmark
(<clearance level>) <+ 1 lookalike set> · <date>
**Benchmark basis:** <level at which the privacy floor cleared; if broadened:
"a privacy feature of the data source, not missing data">
**Freshness:** profile synced <when>; classification <fresh/aging/stale>
Credits used this run: ~<N> (1 profile + 5×<rate calls> + 2 lookalike)

Honesty rules

• Affiliations are corpus-derived, never exhaustive. Section 4's note is mandatory; a clean screen is "no overlap found in our corpus".
• Band ≠ their rate. Section 5 framing is fixed; never imply rate history.
• Placeholder slots stay empty. Never fill section 7 from memory, the web, or estimation — labeled slots only.
• Stale status leads. Freshness classification is in the header, not buried (Freshness Gate).
• PII invariant (convention 7). No contact info appears regardless of source — including anything pulled from other session data sources. Scrub emails/phones/contact links from bio snippets.
• Disclose every Refusal Recovery broaden at the clearance level.

Credit footprint

thorough: ~8–13 credits (1 profile + 1–2 rate calls + 2 lookalike) · thrifty: ~6 credits (1 profile + 1 rate call, max 2 ladder rungs, no comp set)