auto-decorrelation

What problem does it solve?

Single-model AI verification misses critical blind spots because a model cannot identify flaws inherent to its own architecture family, leading to lower-quality checks for high-stakes project changes that other components rely on.

Core Features & Use Cases

• Cross-Family Verifier Recruitment: Automatically discovers and recruits available AI sidecars (Codex, Gemini, local GPU models) with different model families than the governor to perform adversarial verification of load-bearing changes.
• Graceful Degradation: Falls back to same-family in-session verification with honest notes when no cross-family sidecars are available, never hard-failing due to missing tools.
• Use Case: When you are modifying a project's gate infrastructure, onboarding scaffolds, or a skill that produces trusted input for downstream workflows, use this skill to get a decorrelated verification that catches issues a single model would miss.

Quick Start

Use the auto-decorrelation skill to recruit a cross-family verifier for your latest load-bearing project change.