The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the openspec CLI to manage project metadata and documentation. It guides the agent to use subcommands such as status, new change, and instructions while incorporating context-specific identifiers like change names and store IDs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the local repository, including openspec/config.yaml and various markdown artifacts (e.g., spec.md, design.md).
Ingestion points: Reads files from the openspec/ root and artifact paths identified by the CLI.
Boundary markers: Explicitly instructs the agent to treat these files as constraints and prohibits copying their content into the conversation or new artifacts.
Capability inventory: Limited to openspec CLI subcommands via allowed-tools; the skill contains a specific guardrail prohibiting the implementation of application code.
Sanitization: Relies on the underlying CLI and the platform's tool restrictions for safe handling of project data.

openspec-explore