[prompts] extract / score / validate-claims prompt pack

[amirbahador-hub]

Labels: prompts, phase-1
Context: The three prompts a power user's Claude Code agent runs. They must produce output
conforming to EvalResult.
Scope:

• New packages/prompts with extract.md, score.md, validate-claims.md.
• extract.md: resume text → structured Resume.
• score.md: Resume (+ optional JD as keyword context) + rubric + detected archetype → scored
  EvalResult with ≥3 specific issues (quoted text + concrete fix).
• validate-claims.md: flag unsupported claims / tool-or-tech hallucinations → Claim[].
• Each prompt states its exact output JSON shape (referencing the schema) and embeds/links the
  rubric + archetype weights.
  Acceptance:
  • 3 prompt files present, each specifying schema-conformant output
  • A snapshot test asserts each prompt embeds the current rubricVersion
    Out of scope: rewrite/tailor prompts (Phase 2).