The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external media files that could contain malicious instructions in metadata or subtitles, which an agent might misinterpret during analysis.
Ingestion points: Input media files are processed in SKILL.md and reference.md via ffmpeg -i and ffprobe commands.
Boundary markers: Absent. There are no instructions to the agent to ignore or delimit potentially untrusted content within media file metadata.
Capability inventory: The skill performs file writing (creating output media and list files) and command execution via ffmpeg and ffprobe in SKILL.md and reference.md.
Sanitization: Absent. No filtering or sanitization of metadata is performed before the agent processes or displays file information.
[DYNAMIC_EXECUTION]: The skill provides templates for generating shell scripts to automate media export tasks.
Evidence: SKILL.md contains a template for export-all-platforms.sh which uses fixed logic to automate multiple FFmpeg commands.

ffmpeg