The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the Bash(openspec:*) tool to interact with the project. It constructs shell commands using variables such as <name>, <id>, and <artifact-id> which are derived from user input or the project's internal state. This creates a dependency on the CLI tool's own handling of these arguments.
[INDIRECT_PROMPT_INJECTION]: The skill is explicitly designed to read files, search code, and ingest project-specific configuration from openspec/config.yaml. It is instructed to apply context and rules from these files as behavioral constraints. If a project contains malicious instructions within these files or the codebase being explored, it could potentially influence the agent's logic or lead to unintended actions.
[DYNAMIC_EXECUTION]: The instructions specify that if a command's output delegates the creation of an artifact to a specific skill or command, the agent should invoke it. This represents a dynamic execution flow where data returned by the CLI tool at runtime determines which further tools or commands are executed.

openspec-explore