The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read local project configuration files (e.g., config.yaml) and existing OpenSpec artifacts to gain context. While these files are external inputs that could theoretically contain malicious instructions, the skill is explicitly instructed to treat them as constraints rather than instructions to reproduce or execute.
Ingestion points: Reads project context from openspec/config.yaml and existing artifacts from artifactPaths.
Boundary markers: The skill instructions lack explicit delimiters for external content, but include behavioral guardrails against content reproduction.
Capability inventory: Restricted to the openspec CLI tool via Bash(openspec:*) permissions.
Sanitization: The skill relies on agent reasoning and explicit instructions to avoid implementing application code.
[COMMAND_EXECUTION]: The skill uses the openspec CLI for project management and artifact creation. This is a restricted usage defined in the allowed-tools frontmatter, preventing the execution of arbitrary shell commands outside the openspec scope.

openspec-explore