The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted environmental data from Phase I and Phase II ESA reports and interpolates this content into generated markdown reports, creating an indirect prompt injection surface.
Ingestion points: The site_address and various fields within phase_1_esa (findings, recs, data_gaps) and phase_2_esa (exceedances, contaminants) are ingested in environmental_risk_calculator.py and formatted in modules/output_formatters.py.
Boundary markers: The skill does not use explicit boundary markers or instructions to ignore embedded commands when interpolating untrusted data into the reports.
Capability inventory: The skill has the capability to create directories and write files to the local filesystem (found in environmental_risk_calculator.py).
Sanitization: While the site_address is partially sanitized for use in filenames by replacing spaces and slashes, no sanitization is performed on other fields before they are included in the report body.
[DYNAMIC_EXECUTION]: The main script environmental_risk_calculator.py and several of its modules (e.g., cleanup_cost_estimation.py, output_formatters.py) use sys.path.insert to dynamically modify the Python module search path at runtime to resolve local utility dependencies.
[EXTERNAL_DOWNLOADS]: The skill requires several external Python packages for its operations, including pytz for timezone handling, and numpy, numpy_financial, pandas, and scipy for financial and statistical calculations. These are well-known and established libraries in the data science ecosystem.

environmental-due-diligence-expert