What problem does it solve? AI-generated artifacts, workflows, and architecture designs often contain contradictions, hallucinated claims, missing dependencies, or security flaws that slip through self-review. This Skill applies multi-layer validation and returns structured verdicts (VALID / INVALID / UNKNOWN) with confidence scores, so nothing ships on unverified claims. ## Core Features & Use Cases - Multi-Layer Validation: Logical, schema, dependency, orchestration, contradiction, hallucination, architecture, and security checks in one pass. - Confidence-Scored Verdicts: Every check returns VALID, INVALID, or UNKNOWN with a 0.0–1.0 confidence score; UNKNOWN is preferred over guessing. - Claim-Evidence Layer: Factual claims must carry evidence tags ([E]/[C]/[S]) pointing to artifacts that actually exist on disk, with falsification-first testing. - Security & Prompt QA: Detects hardcoded secrets, prompt injection, SQL injection, PII exposure, and validates API contracts and prompt quality. - Use Case: Before deploying an n8n workflow or publishing an AI prompt, ask for a validation pass and receive a structured report listing blockers, warnings, and recommendations. ## Quick Start Ask the assistant to validate your workflow, architecture design, or prompt before deployment and request a structured verdict report with confidence scores.
