The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill utilizes a vee_session token for authentication with the hosted backend at riffkit.ai. It explicitly instructs the agent to treat this token as a secret, forbidding logging, echoing, or persistence of the credential.
[COMMAND_EXECUTION]: The skill performs billable actions via POST /api/riffs. To mitigate financial risk, the instructions mandate explicit user confirmation before submission and prohibit automated retries of failed tasks.
[REMOTE_CODE_EXECUTION]: The skill documentation explicitly forbids fetching and executing instructions from external URLs at runtime, ensuring the agent operates only within the scope of the reviewed instructions.

riffkit