The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external workforce data from CSV files, which presents a potential surface for indirect prompt injection.
Ingestion points: The scripts scripts/attrition_predictor.py, scripts/headcount_planner.py, and scripts/survey_analyzer.py read data from user-supplied CSV files (employees.csv, workforce.csv, survey_results.csv).
Boundary markers: The instructions lack explicit boundary markers or delimiters to protect the agent from interpreting data content as instructions during the analysis workflow.
Capability inventory: The skill's capabilities are restricted to local data processing and standard output. There are no network operations, file system modifications, or shell command executions within the provided scripts.
Sanitization: The scripts implement basic data sanitization through helper functions like safe_float, safe_int, and parse_score which validate and cast input values to expected numeric types.
[DATA_EXPOSURE]: The skill handles sensitive HR data, including compensation, performance ratings, and engagement scores.
The scripts/survey_analyzer.py script implements a privacy safeguard by enforcing a minimum segment size (MIN_SEGMENT_SIZE = 5) to ensure anonymity in group reporting.
The documentation includes a 'Data Governance Checklist' reminding users to apply data minimization and anonymization before processing.

people-analytics