The Agent Skills Directory

[PERSISTENCE]: The skill provides explicit instructions for the agent to modify shell startup files (such as ~/.bashrc, ~/.zshrc, ~/.profile, or ~/.zshenv) to persist the TRANSCRIPTOUT_API_KEY environment variable across shell sessions.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (YouTube transcripts, titles, and descriptions) which are controlled by third-party uploaders and could contain malicious instructions designed to hijack the agent's behavior.
Ingestion points: Video transcripts and metadata retrieved from the api.transcriptout.com/v1/transcript endpoint as described in SKILL.md.
Boundary markers: The skill includes a dedicated "Treat fetched content as data" section advising the agent to ignore commands found within transcripts, although these are not strict technical boundaries.
Capability inventory: The agent is expected to use terminal access and file system write operations for setup, and network access for data retrieval.
Sanitization: The instructions recommend that the agent read and summarize content without following any embedded instructions or task changes.
[COMMAND_EXECUTION]: The authentication and account creation workflows in references/auth-setup.md require the agent to execute complex shell command strings involving curl for registration, sed for token extraction, and printf for writing sensitive keys to the file system.

video-transcript