The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input via the $INPUT variable which is interpolated directly into shell commands throughout the file.
Ingestion points: The $INPUT variable in SKILL.md represents a user-provided file path.
Boundary markers: Absent. The instructions do not include delimiters or warnings to ignore instructions embedded in filenames.
Capability inventory: The skill uses the Bash tool to execute ffmpeg, ffprobe, scenedetect, rm, grep, awk, and tail.
Sanitization: Absent. There is no evidence of path validation or shell escaping for the input variable.
[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to perform complex video inspection tasks. It specifically instructs the agent that these operations are safe to auto-execute without confirmation, which reduces user oversight during execution.
[DYNAMIC_EXECUTION]: The skill includes a 'Optimal CRF Finder' section that uses a bash for loop to dynamically generate and execute multiple encoding commands, file size checks, and quality assessments at runtime. This involves creating temporary files in /tmp/ and executing cleanup commands (rm -f).

claude-video-analyze