The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates via the ocx CLI tool, allowing the agent to manage a local proxy service. This includes performing service restarts (ocx service restart), modifying system-level configurations, and managing local storage archives.
[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for reading and processing request logs (ocx logs) and usage data, which are sourced from external interactions. This presents a potential surface where the agent could ingest malicious instructions embedded in log data.
Ingestion points: Untrusted data enters the agent context via ocx logs, ocx usage, and ocx account list commands (in file SKILL.md and references/01_management_surface.md).
Boundary markers: The skill repeatedly instructs the agent to use the --json flag to ensure it processes structured data rather than raw text, reducing the risk of accidental instruction following.
Capability inventory: The agent possesses significant management capabilities, including ocx account pause, ocx storage cleanup --yes, and ocx service restart.
Sanitization: The instructions emphasize verifying proxy status via ocx ready and ocx status before performing management actions, and it explicitly forbids the agent from handling plaintext credentials returned by CLI commands.

ocx