The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from an external source (YouTube transcripts) which could contain malicious instructions designed to influence the agent's behavior. * Ingestion points: extract_transcript.py (line 33) fetches data using the youtube-transcript-api library. * Boundary markers: Absent. The output format does not wrap the fetched text in delimiters or provide instructions to the agent to ignore embedded commands. * Capability inventory: The skill has file-writing capabilities (line 59) and outputs to stdout, which the agent processes. * Sanitization: Absent. The script joins raw transcript text values without filtering or escaping.
[COMMAND_EXECUTION]: The script takes a user-supplied file path as an argument and writes the transcript content to that location using Path.write_text(). While this is an intended feature for saving transcripts, it enables the agent to write to arbitrary paths on the file system if the agent is not properly constrained by the platform's security environment.

extract-transcript