The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/frame.sh constructs a shell command for ffmpeg by interpolating the $time and $index variables. Because these variables are not sanitized for shell-active characters like backticks or $(), an attacker could potentially execute arbitrary commands if they can control these inputs.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface.\n
Ingestion points: Command-line arguments --time, --index, and the video file path in scripts/frame.sh.\n
Boundary markers: Absent. There are no delimiters or instructions to treat the inputs strictly as data.\n
Capability inventory: Subprocess execution of ffmpeg and directory creation via mkdir in scripts/frame.sh.\n
Sanitization: Basic shell quoting is used in scripts/frame.sh to prevent word-splitting, but it is insufficient to prevent command substitution or other shell-based injections.

video-frames