The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads quantitative market data from the domain pandadata.pandaaiquant.com using the scripts/pandadata_market_data.py script. This domain is managed by the vendor (quantskills) to support the skill's primary function of factor validation.
[COMMAND_EXECUTION]: The scripts/check_dependencies.py script uses subprocess.run to execute pip install for environment setup. This is a standard utility function to ensure the local Python environment has the necessary packages (like pandas, numpy, and panda_data) to run the replication pipeline.
[DYNAMIC_EXECUTION]: The utility script scripts/check_dependencies.py employs importlib.import_module to dynamically verify if required Python packages are installed before proceeding with research tasks.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of quantitative reports, PDFs, and web pages.
Ingestion points: External files provided by the user via the qsh-form task placeholder are processed in 01_translation/ and 02_factor_reproduction/.
Boundary markers: The skill instructions explicitly require the agent to mark uncertain OCR/PDF extraction regions with tags like 待核验 (pending verification).
Capability inventory: The skill can write files, execute localized Python scripts for backtesting (local_backtest.py), and perform network requests for market data.
Sanitization: The skill relies on standard Python libraries (PyMuPDF, pdfplumber) for extraction, and the final output is subject to a quality gate check (quality_gate_check.py).
[CREDENTIALS_SAFE]: The skill follows security best practices by instructing the agent to read credentials from environment variables or a local .env file (~/.pandadata/pandadata.env) rather than hardcoding secrets in code or reports.

report-replication