The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to interact with the system using the circuit CLI tool. This includes commands for running workflows, checking system readiness (circuit doctor), and managing project configuration.- [INDIRECT_PROMPT_INJECTION]: The skill addresses the risk of processing untrusted user input by providing specific guidelines for shell-safe command construction.
Ingestion points: The --goal argument in the circuit run command, which accepts arbitrary task descriptions from the user.
Boundary markers: The instructions mandate the use of single quotes for goal text and provide escaping procedures for apostrophes.
Capability inventory: Includes the ability to modify project files (via fix and build flows), execute shell commands, and access external LLM APIs through configured connectors.
Sanitization: Guidance is provided for manual escaping of user-provided strings before they are passed to the shell.- [EXTERNAL_DOWNLOADS]: The skill mentions external dependencies such as the @openai/codex Node.js package. These references are to well-known, official tools from established service providers and are used for their intended purpose of providing model access.

circuit