The Agent Skills Directory

[COMMAND_EXECUTION]: The testing scripts scripts/test_interview_state.py and scripts/test_visualize_interview.py utilize subprocess.run to verify the command-line interface of the skill's utility scripts. These calls are confined to local regression testing and do not process untrusted external input.
[EXTERNAL_DOWNLOADS]: The README.md documentation describes installing the skill via npx from the author's GitHub repository. It also references official guidelines from well-known services like Cloudflare and Vercel Labs in illustrative examples, following standard development practices.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input to generate requirements. It mitigates injection risks by implementing separate, read-only worker contracts in references/evaluator.md and references/questioner.md that explicitly instruct the agent to treat data as evidence rather than instructions.
[DYNAMIC_EXECUTION]: The skill's Python test suite dynamically loads local modules using importlib.util for unit testing purposes. This behavior is restricted to the skill's internal files and is a standard practice for modular script development.

interview