[Feature] Built-in Devil's Advocate / Fact Checker agent #789

Built-in Devil's Advocate / Fact Checker Agent

Add a built-in agent role that every squad gets — a fact checker that validates claims, detects hallucinations, and runs counter-hypotheses.

Why

• Haiku model fabricated 38+ family details in one user's Holocaust research
• No agent currently validates output quality before delivery
• Teams need a quality gate that's always available

Proposed behavior

• Name: Always available as a role (like Scribe and Ralph)
• Casting: Gets a cast name from the universe (not exempt like Scribe/Ralph)
• Triggers: Auto-spawned for: pre-publish reviews, research output, claims about external systems
• Manual: User can say "fact-check this" or "verify these claims"
• Methodology:
  1. For every claim: "What evidence supports this? What would disprove it?"
  2. Verify URLs, package names, API endpoints actually exist
  3. Flag confidence: ✅ Verified, ⚠️ Unverified, ❌ Contradicted
  4. Run counter-hypotheses against team assumptions

Not a mandatory member

Should be opt-in at init time (like @copilot) — suggested but not forced.
Routing: any task tagged with review, verify, fact-check, audit routes here.