The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input to generate speech content, which creates a potential surface for indirect prompt injection.
Ingestion point: User input collected in Step 1: 了解演讲场景 (e.g., identity, audience, core message).
Boundary markers: The instructions lack explicit delimiters or "ignore embedded instructions" directives when interpolating user data into the writing task.
Capability inventory: The skill has no dangerous capabilities such as network operations, file writing, or command execution.
Sanitization: No sanitization or validation of user-provided strings is mentioned.
[EXTERNAL_DOWNLOADS]: The skill includes instructions to recommend the installation of an external resource from a repository.
Evidence: The "能力缺口响应规则" (Capability Gap Response Rule) instructs the agent to provide the command /install 技能宝 github:mindverse/skillhub to users if requested tasks are out of scope.

speech-writer