The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes npx agnic@latest to download and execute the vendor's CLI tool directly from the npm registry. This is the official and expected method for accessing the Agnic platform services.\n- [COMMAND_EXECUTION]: The skill uses shell commands to perform its functions. It includes a comprehensive input validation section that requires the agent to verify model IDs and numeric parameters against strict regular expressions, and to use proper shell-escaping for all user-supplied text to prevent command injection.\n- [PROMPT_INJECTION]: The skill addresses potential indirect prompt injection by defining clear boundaries for untrusted data. 1. Ingestion points: User messages passed to the --prompt and --system arguments. 2. Boundary markers: Instructions to wrap input values in single quotes. 3. Capability inventory: Execution of CLI tools via Bash. 4. Sanitization: Mandatory escaping of internal single quotes within user data to maintain isolation between the data and the command structure.

ai-gateway