The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various local system commands to process video files and fingerprint the environment.
It utilizes ffmpeg, ffprobe, and whisper for media processing.
It creates a dedicated temporary directory with restricted permissions (chmod 700) to store intermediate frames and audio.
It performs environment fingerprinting by checking the OS version, shell type, and installed development tools (Python, Node, Homebrew).
It includes a mandatory cleanup step to remove temporary files after processing.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from external video recordings.
Ingestion points: The skill reads a user-provided video file path and processes audio transcriptions generated from that file.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard potentially malicious commands embedded within the video narration or text appearing in the frames.
Capability inventory: The skill has the capability to run system commands, fingerprint the environment, and write/execute new automation scripts (Bash, Python, Node.js).
Sanitization: The skill lacks mechanisms to sanitize or validate the content of the transcriptions before the agent uses them to reconstruct the workflow and propose automation code.

automate-this