The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The workflow involves processing untrusted external data, such as transcripts of recordings, to automate editing decisions and code generation. This creates a surface where instructions hidden in the transcript could potentially influence the agent's behavior.
Ingestion points: Processing of raw recordings and transcripts in Layer 2 (Organization) and highlight extraction sections of SKILL.md.
Boundary markers: Example prompts lack specific markers or instructions to isolate the data being analyzed from the agent's operating instructions.
Capability inventory: The skill utilizes FFmpeg for file processing and the requests library for external API interactions in SKILL.md.
Sanitization: The skill does not describe any sanitization or validation steps for the content of the transcripts before they are processed by the agent.
[DATA_EXFILTRATION]: The skill performs network requests to external services to generate audio assets.
Evidence: The Python example in SKILL.md makes POST requests to the ElevenLabs API at https://api.elevenlabs.io/v1/text-to-speech/ to generate narration.
[COMMAND_EXECUTION]: The skill generates and executes shell commands for technical media tasks.
Evidence: SKILL.md contains numerous examples of FFmpeg commands and shell loops for trimming, concatenating, and extracting data from video files.

video-editing