The Agent Skills Directory

[SAFE]: The skill provides instructions for a multi-agent validation workflow intended to reduce errors and bias. It emphasizes critical evaluation of all reviewer output, requires the agent to adjudicate rather than count votes, and warns against blindly trusting external models. No patterns of obfuscation, malicious command execution, or unauthorized data access were detected. The use of external model families via CLI or API is described as a high-level capability for verification and does not include suspicious implementation details or hardcoded credentials.

second-opinion