The Agent Skills Directory

[SAFE]: The skill implements a robust lateral thinking framework using personas. It includes clear instructions for argument parsing, context gathering, and response synthesis.
[SAFE]: It utilizes a specific MCP tool (ouroboros_lateral_think) for its core functionality and follows a 'deferred-schema' pattern for tool discovery, which is a standard pattern for this platform.
[SAFE]: The skill handles data ingestion by explicitly asking the user for missing context (ask_user) rather than fabricating data, which is a security best practice.
[SAFE]: While it mentions a Base64-encoded dispatch block, analysis of the protocol shows it is used for structured data transport between the tool and the host to coordinate sub-agents, not for obfuscating malicious instructions.

unstuck