cross-national

Pass

Audited by Gen Agent Trust Hub on Oct 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external survey datasets in CSV format (KNHANES, NHANES, CHNS), which constitutes an attack surface for indirect prompt injection.
  • Ingestion points: The skill loads data from single and multiple CSV files from user-defined paths as described in Phase 2 and the CHNS Reference section.
  • Boundary markers: The instructions do not specify the use of boundary markers or safety delimiters to isolate the agent from instructions potentially hidden within the dataset values.
  • Capability inventory: The agent has access to Bash (used to execute R scripts), Write, and Edit tools, allowing it to act on instructions.
  • Sanitization: No specific sanitization, validation, or filtering of the CSV content is required before the data is used to generate analysis scripts.
  • [DYNAMIC_EXECUTION]: The skill involves the generation and execution of R analysis scripts based on the provided data.
  • Evidence: The workflow explicitly outputs and executes analysis_korea.R and analysis_us.R as self-contained analysis scripts.
  • Mechanism: The agent maps variables and generates R code to perform weighted survey analysis; while this is a primary function of the skill, the generation of executable code from external data sources is a known dynamic execution pattern.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 4, 2026, 01:27 PM