Rfp Response Builder

RFP Response Builder

Holding-company and agency strategy leads answer RFPs, and the creator-strategy section needs to look institutional: market sizing, example slates per market, indicative rate bands — all cited. This skill reads an RFP's requirements, treats each target market/archetype as its own search (convention 9: a target list becomes a slate, never one merged search), and assembles a per-market, citation-ready strategy section. Heavy multi-call — the credit estimate always fires.

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 RFP is really a single written brief for one market, brief-to-shortlist is lighter. This skill is for multi-market / multi-archetype RFPs that need market sizing + cited bands as a formal response section.

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 RFP (required) — pasted text or excerpt. Normalize it yourself; do not make the user restructure it.
• Target markets / archetypes — extract the LIST from the RFP (e.g. "US + UK + DE beauty mid-tier", "3 creator archetypes"). Confirm the list back in one block. This list defines the per-target sections.
• Example-slate size per market — default 5; honor "a few"/"top 10".
• Deal types / compensation model if the RFP specifies one.
• Credit mode — default thorough; thrifty on the standard triggers.

Flow

Step 0 — Normalize + enumerate targets (no tool call). Produce the target matrix: market × archetype × tier, with the RFP's own requirement language per cell. Show it; proceed unless corrected. Convention 9: one search per target — never one merged generic search.

Pre-run estimate fires here, always (this skill is multi-call by nature). For T targets at slate size S, thorough ≈ T×(2 search + S profiles) + T market-intel calls. Example: 3 markets × (2 + 5) + 3×5 = 36 credits. State it and offer thrifty before fanning out.

Step 1 — Market sizing per target — query_market_intelligence { "mode": "market", "vertical": "<target vertical>", "sub_category": "<if RFP specifies>", "company_type": "<if relevant>" } — wrapped in Refusal Recovery (market floor 5 brands / 25 deals; thorough: full ladder; thrifty: max 2 rungs). Gives deal volume / distribution to size the opportunity per market. Disclose clearance level. 5 credits each incl. ladder retries.

Step 2 — Example slate per target — one search per target: search_creators { "mode": "brief", "brief": "<normalized per-target requirement text, ≤2000 chars>", "filters": { "country": "<market>", "platform": "<if stated>", "min_followers": <tier floor>, "max_followers": <tier ceiling>, "niche": "<if clean>" }, "limit": <2×slate size>, "precision": "balanced" } 2 credits per target. Thin-niche honesty (convention 8): if a target's niche is thin in-corpus, disclose and broaden exactly one labeled step at a time (exact niche → adjacent → vertical), marking each pick exact vs adjacent fit. Cross-market gaps (RFP constraint): a market where even the broadened search returns thin is flagged "needs local sourcing" — never padded.

Step 3 — Profile the slate (thorough) — get_creator_profile per finalist { "identifier": { "type": "creatorland_user_id", "creatorland_user_id": "<id>" } } (or whichever single type the search returned) → audience-geo confirmation + freshness. Thorough: profile each target's slate cut. Thrifty: skip profiling, present slates on search signal only and say so. 1 credit each.

Step 4 — Freshness Gate per profiled creator; stale → "re-verify before pitch" per market.

Step 5 — Indicative rate band per target — reuse the rate dimension: query_market_intelligence { "mode": "rate", "vertical": "<target vertical>", "deal_type": "<if specified>", "creator_tier": "<the target's slate tier, if it clusters at one>" } wrapped in Refusal Recovery (rate floor 10 brands / 50 deals). One band per distinct vertical/market — cited. When the target's example 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-level band as the fallback and say so (a too-thin tier auto-broadens, disclosed). 5 credits each.

Step 6 — Assemble (no tool calls) the per-market strategy document.

Deliverable

A per-market creator-strategy section in markdown, citation-ready:

# Creator Strategy — <RFP / brand name> response
_Prepared with Creatorland Data · <date> · <T> target markets_

## Approach & methodology
Per the RFP we treated each target market/archetype as its own search.
Market sizing draws on the Creatorland deal corpus; example slates are
illustrative (not a committed roster); rate bands are corpus-level vertical
benchmarks. Privacy floors mean some narrow slices were widened — disclosed
inline per market.

## Market: <market 1> — <archetype/tier>
**Opportunity size:** <deal volume / distribution from market mode>
> <provenance line verbatim> · recency window: <window>
**Example talent slate** (illustrative, <S> creators):
- @<h> (<platform>, <followers>) — <fit note>; audience <geo %>; freshness
  <fresh/aging>; <exact fit | adjacent fit — niche>
<repeat per creator>
**Indicative rate band:** market band for **<vertical>** —
p25 $<x> · median $<y> · p75 $<z> (corpus-level, not per-creator).
> <provenance line verbatim>
**Coverage note:** <"exact-niche depth good" | "N exact + M adjacent" |
"⚠ thin in corpus — recommend local sourcing partner for this market">
**Benchmark basis:** <clearance level; if broadened: privacy-floor note>

## Market: <market 2> ...
<repeat the block per target>

## Cross-market summary
| Market | Corpus depth | Slate confidence | Rate band basis |
|---|---|---|---|
<one row per market — honest depth/confidence, gaps flagged>

## Caveats
- Slates are illustrative, not committed rosters.
- Rate bands are vertical-level corpus benchmarks — no per-creator rates exist.
- Conflict/affiliation screening available on request (pro plan) — see
  `conflict-check`.
- No creator contact info is included or available via this tool.

---
Data freshness: <N>/<M> profiled creators synced within the last sync window;
<K> flagged for re-verification.
Provenance: Creatorland Data MCP · <searches> + <profiles> + <market-intel
calls> · <date>.
Credits used this run: ~<N> (<breakdown by call type>).

Honesty rules

• Each target is its own search (convention 9). Never merge archetypes into one generic search; the document is per-target sectioned.
• Thin markets are flagged, never padded (conventions 8 + RFP constraint). Disclose, broaden one labeled step at a time, mark exact vs adjacent fit; a market still thin after broadening is "needs local sourcing".
• Slates are illustrative. Never present them as committed rosters or imply availability.
• Rate bands are vertical-level. Cited corpus band, never per-creator rate.
• Disclose every Refusal Recovery broaden at the clearance level, per market.
• PII invariant (convention 7). No contact info, regardless of source.

Credit footprint

thorough: ~(T×(2 + S) + 2T×5) credits — e.g. 3 markets × slate 5 ≈ 51 credits (3 searches + 15 profiles + 3 market + 3 rate calls; +5 per ladder rung) · thrifty: ~(T×2 + T×5) credits — e.g. ~21 (searches + one market-or-rate band per market, no profiling, max 2 ladder rungs). Estimate stated up front, always.