The Agent Skills Directory

[PROMPT_INJECTION]: The skill's setup documentation in references/auth-setup.md instructs the agent to intentionally circumvent platform-level security redaction filters. It directs the agent to write raw API responses containing sensitive credentials (like access tokens and API keys) to temporary files and read them back for subsequent requests, specifically to ensure these values are not hidden or masked by the environment's security protocols that monitor tool output.
[PERSISTENCE]: The skill requires the agent to persistently store the TRANSCRIPT_API_KEY across sessions. The instructions in references/auth-setup.md suggest modifying system-level configuration files such as shell profiles (e.g., .bashrc, .zshrc, or .profile), which establishes a persistence mechanism for environment variables on the host system.
[INDIRECT_PROMPT_INJECTION]: The skill processes YouTube transcript data, which is untrusted external content and could contain malicious instructions designed to influence the agent's behavior during summarization or analysis.
Ingestion points: Fetches external transcript text from transcriptapi.com via GET requests (File: SKILL.md).
Boundary markers: No specific delimiters or "ignore instructions" warnings are provided to wrap the ingested transcript text.
Capability inventory: The skill possesses the ability to make network requests and modify local configuration files (File: SKILL.md, references/auth-setup.md).
Sanitization: The skill does not describe any sanitization, validation, or filtering of the fetched transcript before it is passed to the agent's context.

transcript