The Agent Skills Directory

[PERSISTENCE]: The skill's setup guide in references/auth-setup.md instructs the agent to persist the TRANSCRIPT_API_KEY by modifying shell configuration files such as ~/.bashrc, ~/.zshrc, or other environment-specific profiles. This is intended to ensure the key is available across sessions.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process transcripts and metadata from YouTube videos via the transcriptapi.com service. External video content is untrusted and could contain text designed to influence or hijack the agent's behavior.
Ingestion points: https://transcriptapi.com/api/v2/youtube/transcript (referenced in SKILL.md and references/curl-examples.md).
Boundary markers: The instructions do not define specific delimiters or "ignore" instructions for the agent when processing transcript text.
Capability inventory: The skill uses network GET requests to retrieve content; it does not perform file writes or command execution based on the retrieved data.
Sanitization: No specific sanitization or filtering of transcript content is described.
[CREDENTIALS_UNSAFE]: The references/auth-setup.md file contains explicit instructions for the agent to bypass platform-level security redaction of secrets. It suggests writing raw HTTP responses (containing access_token and api_key) to temporary files to prevent the values from being hidden or redacted by the agent environment's output filters.

youtube-full