The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references and installs the lunar_python and tzdata packages from standard registries. These are well-known, legitimate libraries used for lunar calendar calculations and timezone database information.
[COMMAND_EXECUTION]: The skill executes a bundled local script (scripts/qimen_cli.py) to handle complex metaphysical calculations. This ensures deterministic results and prevents the AI from hallucinating complex astronomical data. The script operates entirely on local JSON input/output and does not perform network operations or access sensitive system files.
[INDIRECT_PROMPT_INJECTION]: The skill collects user-supplied descriptions of their questions and locations, which creates an ingestion surface for untrusted data.
Ingestion points: User responses provided during the 'Step 1: Interview' phase described in SKILL.md.
Boundary markers: The agent is instructed to place user input into a structured JSON file (tmp/qimen_input.json) before calling the calculation script.
Capability inventory: The skill uses a subprocess call to run the Python script and then reads the script's JSON output to generate a final response.
Sanitization: The qimen_cli.py script utilizes strict JSON parsing (json.loads) and maps inputs to predefined data structures and types, which effectively sanitizes the data before it influences the agent's interpretation.

qimen-dunjia