The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input and interpolates it into prompts for secondary agents, creating a potential vector for indirect prompt injection. * Ingestion points: The {用户的话题} (user topic) variable in SKILL.md is populated by user input. * Boundary markers: There are no explicit delimiters or instructions to the sub-agents to ignore potentially malicious commands within the user-provided topic. * Capability inventory: The skill uses an Agent tool to call and execute multiple independent AI agents. * Sanitization: The skill lacks validation or escaping mechanisms for the user input before it is passed to the Agent tool.
[DYNAMIC_EXECUTION]: The skill dynamically constructs prompts for multiple independent agents based on user-selected expert personas. * Evidence: SKILL.md contains a structured prompt template that is populated at runtime with expert-specific thinking methods and the user's topic to drive the behavior of sub-agents.

targeted-chatroom