The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the openspec CLI tool to query the status of active project changes via openspec list --json. This is a legitimate functional requirement using the vendor's own infrastructure.
[DATA_EXPOSURE]: The agent is instructed to read local codebase files and OpenSpec-specific markdown artifacts (such as proposals, designs, and specs) to gain context. This data ingestion is restricted to the agent's reasoning process and no exfiltration patterns were detected.
[PROMPT_INJECTION]: Instructional blocks such as "IMPORTANT: Explore mode is for thinking, not implementing" and "NEVER write code" are used to define the skill's operational boundaries. These commands act as functional constraints rather than attempts to bypass security filters.
[SAFE]: No malicious obfuscation, remote code execution from unknown sources, privilege escalation, or persistence mechanisms were found in the skill definitions.

openspec-explore