The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests user-supplied divination questions that are subsequently interpolated into an interpretation prompt for the AI agent. This surface is well-secured: it includes a deterministic safety classifier in scripts/classify_sensitive.py that redirects high-risk queries (e.g., medical, legal, or psychological crises) to professional help, and it employs a fact-auditing mechanism in scripts/verify_facts.py to ensure the agent's interpretation does not fabricate or alter the core hexagram data.\n- [COMMAND_EXECUTION]: The skill's workflow involves the agent executing local Python scripts to perform divination charting and calendar conversion. These commands are restricted to the scripts bundled with the skill and are necessary for its intended purpose of providing a deterministic divination engine.\n- [REMOTE_CODE_EXECUTION]: Analysis of the Python scripts (specifically scripts/run_liuyao.py) revealed the use of importlib.import_module. This is used for a self-check procedure to verify that the vendored lunar_python library is being loaded from the correct local path. No pattern of loading external or user-controlled code was found.

fortune-liuyao