The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute a CLI tool named siyu-plan when the runtime environment allows. This execution is protected by a SHA256 integrity check of the route-contract.json file, which is a proactive security measure to ensure the routing logic has not been tampered with.
[EXTERNAL_DOWNLOADS]: The skill links to external resources on GitHub and Feishu for documentation and user feedback. These are well-known services and the resources originate from the author's own verified infrastructure.
[PROMPT_INJECTION]: The skill ingests user queries through the /siyu command to determine the appropriate sub-skill for routing. This represents an indirect prompt injection surface. The mandatory evidence chain includes: 1) Ingestion point: User queries in SKILL.md. 2) Boundary markers: Defined interaction modes (Newbie, Routing, Navigation). 3) Capabilities: CLI execution and sub-skill invocation. 4) Sanitization: SHA256 integrity verification of the routing contract.

majia-siyu