The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains explicit instructions to circumvent platform security controls designed to mask sensitive data. It directs the agent to hide API keys and access tokens by writing them to temporary files specifically to prevent platform-level redaction filters from detecting and masking the secrets in tool output.
[PERSISTENCE]: The skill instructs the agent to modify system-level environment configuration files, such as shell profiles like .bashrc or .zshrc, to store the TRANSCRIPT_API_KEY permanently. This ensures the credentials persist across sessions and shell environments.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to the external domain transcriptapi.com for account registration, email verification, and YouTube data retrieval. This domain is not recognized as a trusted or well-known service.
[COMMAND_EXECUTION]: The skill provides curl command templates and sequences for the agent to execute network requests via the system shell, including steps to handle authentication tokens and registration flows.
[CREDENTIALS_UNSAFE]: The skill facilitates the acquisition and persistent storage of an API key (prefixed with sk_). The instructions suggest storing this secret in plaintext within shell configuration files, which can lead to accidental exposure if these files are logged, backed up, or shared.

youtube-api