The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides shell command templates for technical video inspection that directly interpolate the $VIDEO variable, which is sourced from user-provided paths or URLs.\n
Evidence: In SKILL.md, the workflow for technical inspection suggests executing ffprobe -v error -show_entries format=duration,size:stream=width,height,r_frame_rate -of json "$VIDEO" and ffmpeg -y -i "$VIDEO" -vf fps=1 /tmp/video-frames/frame-%03d.jpg.\n
Risk: If the agent environment executes these commands in a shell without proper escaping, a malicious filename containing characters like backticks or semicolons could result in arbitrary command execution on the host system.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external media files and URLs, creating a surface for indirect prompt injection attacks.\n
Ingestion points: The skill accepts local paths, uploaded files, and URLs provided by users as source videos for analysis (SKILL.md, Step 1).\n
Boundary markers: There are no instructions or delimiters provided to the agent to isolate or ignore instructions that may be embedded in video metadata, subtitles, or content.\n
Capability inventory: The skill utilizes powerful system tools including ffprobe, ffmpeg, and file system commands like mkdir (SKILL.md, Step 2).\n
Sanitization: The instructions lack any requirement for the agent to validate or sanitize input paths, filenames, or the content of the media before processing it with system utilities.

video-to-superprompt