The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to check for dependencies (ffmpeg, whisper), fingerprint the user's operating environment (OS, shell, installed tools), and process media files.
[DYNAMIC_EXECUTION]: The skill's primary purpose is to generate and execute custom automation scripts (Bash, Python, Node.js, AppleScript) based on user-provided recordings. In Phase 5, the agent is instructed to write these scripts to files and execute them to verify functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from user-provided video files (visual frames and audio transcripts) and uses this data to reconstruct workflows and generate executable code.
Ingestion points: Frames and audio transcripts extracted from video files via ffmpeg and whisper in Phase 1.
Boundary markers: None. The agent is not instructed to isolate or ignore instructions that might be embedded in the transcript or visual content of the recording.
Capability inventory: The skill has broad capabilities including shell command execution, file system writes, and the execution of generated scripts across multiple languages.
Sanitization: None. The extracted transcript is read directly into the session context and used to influence automation proposals.
[PERSISTENCE]: In Phase 4, the skill explicitly proposes creating scheduled tasks using cron or launchd as a "Tier 3" automation option, which establishes persistence on the host system.

automate-this