The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute the wpsnote-cli command-line tool to generate images based on prompts derived from user-provided text. While identified as a resource from the skill's author, executing CLI commands using content influenced by untrusted external text presents a potential injection surface if malicious content is included in the source article.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of user-pasted articles to generate video scripts and image prompts. ● Ingestion points: User-provided long-form articles. ● Boundary markers: No delimiters or instructions are used to separate untrusted content from the transformation logic. ● Capability inventory: The skill can write to and modify WPS Notes and execute shell commands via wpsnote-cli. ● Sanitization: There is no validation or filtering of the input text or the prompts generated for the image tool.- [PROMPT_INJECTION]: The skill defines 'Iron Rules' that explicitly forbid the agent from showing generated content in the chat, forcing it to be written directly to external storage (WPS Notes). This creates a concealment pattern that reduces a user's ability to review and verify the AI's output for malicious content or instructions before it is persisted outside the conversation environment.

short-video-copywriter