The Agent Skills Directory

[COMMAND_EXECUTION]: The skill involves the execution of shell commands (ffmpeg and ffprobe) to process media files. This is the primary intended function of the skill. The instructions mitigate execution risks by mandating a confirmation step (AskUserQuestion) before any processing command is run and requiring the agent to show the full command in 'Pro mode' for user review.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it processes metadata from external media files.
Ingestion points: Media file metadata extracted via ffprobe and formatted as JSON (Step 1).
Boundary markers: The interaction flow requires a human-in-the-loop confirmation (AskUserQuestion) before acting on the ingested data.
Capability inventory: The agent has shell access to run ffmpeg and ffprobe commands.
Sanitization: The skill instructs the agent to request specific metadata fields rather than processing raw, unparsed text, which reduces the potential for malicious injection through metadata fields.

media-processing