The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local binary located at ~/GitHub/scripts/bin/upload-media using subprocess.run to handle media uploads.
Evidence: Found in scripts/media_toolkit_lib/io.py within the upload_local_file function.
[DATA_EXFILTRATION]: The skill accesses a sensitive local file path ~/.secrets/aipodcasting/env to retrieve the backend API key required for its operations.
Evidence: Found in scripts/media_toolkit_lib/api.py in the get_backend_api_key function. This is a documented method for credential management for this tool.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it transcribes user-provided media and returns the resulting text to the agent without sanitization.
Ingestion points: The fetch_text method in scripts/media_toolkit_lib/api.py retrieves transcript data from a URL.
Boundary markers: No explicit boundary markers or 'ignore' instructions are added to the returned transcript text.
Capability inventory: The skill has the ability to execute local commands via subprocess.run and write files to disk.
Sanitization: No sanitization or filtering is performed on the transcription text before it is returned to the agent context.

media-toolkit