icp-scoringlisted

# ICP Scoring A BDR, RevOps operator, or marketing lead with a raw prospect list today has two options: run them all (waste sender reputation + enrichment credits on poor-fit prospects) or hand-filter (slow, inconsistent across operators). This skill produces a third option — a parallel, criteria-grounded score per record with reasoning, split into qualified vs disqualified outputs the downstream campaign skill can consume directly. The skill is **dual-mode**: standalone invocation defaults to `score_0_100` (Sonnet, 4-bucket reasoning); when delegated from `tam-mapping` Phase 7 it switches to `abc` (Haiku, letter-only tier-A/B/C) for cost-efficient TAM-scale runs. **Distinct from `lead-routing` (BC-2725):** this skill is pre-outreach prioritization of raw prospects, not post-reply SF MQL assignment. --- ## Before Starting **Check for product marketing context first.** If `docs/marketing-context.md` exists, read it for entity-specific ICP criteria. If the file does not exist, fall through to the missing-file fallback below — **do not proceed with degraded inferred ICPs.** ICP criteria are the foundation of every score; running without them produces noise. ### Entity detection The skill scores against ONE Brite entity's ICP per invocation (Nites residential / Supply installer / Labs venue). Detection logic when `--client` is not passed: | State | Behavior | |---|---| | `marketing-context.md` exists with one entity's ICP populated | Use it. Print: `Using entity=<X> from m