The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run shell commands to analyze video files using a variable $VIDEO that is populated by user-supplied paths or URLs. This pattern is vulnerable to command injection if the input contains shell metacharacters.
Evidence: SKILL.md contains the following shell block: ffprobe -v error -show_entries format=duration,size:stream=width,height,r_frame_rate -of json "$VIDEO" and ffmpeg -y -i "$VIDEO" -vf fps=1 /tmp/video-frames/frame-%03d.jpg.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources and incorporates it into agent reasoning without adequate protection.
Ingestion points: Processes local file paths, uploaded files, and external URLs provided by the user (SKILL.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the analysis workflow.
Capability inventory: The skill utilizes shell execution (ffprobe, ffmpeg) and file system access (mkdir, ffmpeg output to /tmp).
Sanitization: There is no evidence of filename sanitization or URL validation before passing these inputs to system commands.
[DYNAMIC_EXECUTION]: The skill utilizes subprocess calls to external binaries at runtime to analyze data.
Evidence: The use of ffprobe and ffmpeg within the SKILL.md workflow represents dynamic execution of system utilities to process content generated or provided during the session.

video-to-superprompt