The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the openspec command-line interface to list and retrieve project metadata (openspec list --json). This is a legitimate use of a vendor tool to establish context within the development environment.
[PROMPT_INJECTION]: The skill includes instructions to the agent to adopt a collaborative 'thinking' posture. It specifically sets boundaries by instructing the agent to 'NUNCA' (never) write code or implement features while in this mode, which serves as a safety guardrail.
[DATA_EXFILTRATION]: While the skill reads local project files for architectural mapping and investigation, it contains no instructions for network operations or external data transfer.

openspec-explore