using-superpowers

The document is a high-risk workflow policy that mandates invoking and following external 'skills' before any agent response. While it does not contain direct malware or hardcoded secrets, it creates a supply-chain and autonomy-abuse vector: forced blind loading of external skill content, prohibition on inspection, and lack of provenance/consent controls materially increase the chance that a malicious or compromised skill will be executed and could exfiltrate data, misuse credentials, or perform harmful actions. Recommended mitigations: require explicit human approval for untrusted skills, ban automatic credential forwarding, add cryptographic signature verification and an allowlist for skills, permit inspection/sandboxing of skill files before execution, and reintroduce the ability to ask clarifying questions prior to invoking external content.