The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external media data, such as transcripts and subtitles, which could potentially contain embedded instructions intended to influence the agent's behavior.
Ingestion points: Materials described in SKILL.md include audio, video, transcripts, subtitles, A-roll, and B-roll.
Boundary markers: The skill does not define explicit delimiters or warnings to isolate processed data from control instructions.
Capability inventory: The skill executes CLI commands (vendor-owned postplus) and routes data to other specialized skills like audio-transcription and editing-decision-engine.
Sanitization: There are no instructions for sanitizing or validating the content of the media files before processing.
[COMMAND_EXECUTION]: The skill instructions allow for the execution of command-line tools and scripts to fulfill routing and confirmation tasks.
Evidence: Mentions direct execution of "owned CLI or script commands" and specifically references the postplus quote confirm command.
Vendor Context: These commands are associated with the skill author's own infrastructure (postplusai).

media-router