Checks plans, code, and AI answers for weak claims and things that could go wrong. Shows the evidence and suggests a test or fix
Fable 5.1 OracleThinks the build through and checks the work, but never writes the code itself
PR ReviewerReviews pull requests risk-first
CASEQuietly reviews your in-progress build from the real artifact and flags weak implementation and design points
Multi-model consensusRuns Claude Code, Codex, and Grok on the same hard question until they agree