HaluCatch / 捕幻

分享海报

@user_48539314/halucatch

Evaluates the reliability of AI Skill execution. Assesses whether a Skill's output is trustworthy, reproducible, and withstands business scrutiny when executed by an AI agent. Covers four dimensions: data pipeline integrity, code risk, business logic ambiguity, and interpretation guardrails. Use when auditing an AI Skill, checking for hallucinations or unreliable outputs, verifying execution reproducibility, or reviewing a Skill's safety before deployment or sharing.