The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to deceive audiences by concealing the commercial nature of sponsored content ("不要在脚本中透露商单性质"). This directive bypasses standard disclosure expectations for advertising.
[COMMAND_EXECUTION]: The skill defines a file-saving operation to a local path (01-内容生产/待发布的内容/中短视频/[标题].md) using a user-supplied title. The absence of sanitization instructions for the title variable could lead to path traversal if the platform does not provide built-in protection.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface. 1. Ingestion points: User-supplied topics and script requests (SKILL.md). 2. Boundary markers: Absent; no delimiters are used to separate user data from system instructions. 3. Capability inventory: File system write capabilities (SKILL.md). 4. Sanitization: No evidence of input validation or escaping for external content processed by the agent.

li-writer