The Agent Skills Directory

[SAFE]: The skill is a template for creating automated verification tasks. Its behavior is consistent with its stated purpose of assisting developers in testing their own code.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (e.g., launch, doctor, drive) that are derived from the project's own environment and configuration. This is a core feature for verifying the application's behavior.
[INDIRECT_PROMPT_INJECTION]: The skill reads and processes the project's repository content to generate the verification skill.
Ingestion points: Codebase analysis during the "Interview the repo" phase (SKILL.md).
Boundary markers: Not explicitly defined in the generation instructions.
Capability inventory: Shell command execution for application launch and testing (SKILL.md).
Sanitization: The skill relies on the agent's internal analysis of the repo rather than executing un-sanitized external strings.
[DYNAMIC_EXECUTION]: The skill generates a verification skill (SKILL.md and features map) and then executes it to confirm functionality. This self-verification step is intended to ensure the generated tool is useful to the developer.

create-verification-skill