The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill generates Python scripts at runtime by interpolating user-provided text into a template and executes these scripts using a specific virtual environment on the host machine. This dynamic code generation and execution process is a high-risk pattern as it lacks robust validation of the injected content.- [COMMAND_EXECUTION]: The workflow relies heavily on executing shell commands via FFmpeg and the Python interpreter. It utilizes user-supplied file paths and generated script paths directly in shell executions, which could be exploited for command injection if paths or script names are maliciously crafted.- [PROMPT_INJECTION]: The skill exhibits a susceptibility to indirect prompt injection through its ingestion of external Markdown scripts. 1. Ingestion points: The skill reads user-provided Markdown scripts for voiceover text and video file paths. 2. Boundary markers: There are no boundary markers or instructions provided to the agent to treat the script content as untrusted data or to ignore embedded instructions. 3. Capability inventory: The skill has extensive capabilities, including writing files, executing shell commands (FFmpeg), and running dynamically generated Python code via subprocess. 4. Sanitization: There is no evidence of input sanitization to prevent characters from escaping the string literals used in the generated Python script or the shell command arguments.

video-voiceover