experiment-design

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests external data by parsing Qualtrics (.qsf) files to extract survey structures and question text. This creates a surface for indirect prompt injection where instructions could be embedded in the survey data.
  • Ingestion points: The Survey mode (Entry Point A) parses user-provided JSON (.qsf) files as described in SKILL.md.
  • Boundary markers: The instructions do not define specific delimiters or warnings for the agent to ignore instructions embedded within the question text.
  • Capability inventory: The agent has the ability to execute generated Bash, R, and Python scripts, write to the filesystem, and ask the user questions.
  • Sanitization: No specific sanitization or filtering logic is described in references/qsf-parsing-guide.md for the extracted survey content.
  • [DYNAMIC_EXECUTION]: The skill generates and executes R and Python scripts for power analysis and Monte Carlo simulations.
  • Evidence: The Power mode workflow in SKILL.md involves generating scripts and saving them to the project directory (e.g., code/power_analysis.R) followed by execution to produce results.
  • Security Context: This functionality is the primary purpose of the skill. The generated scripts use established statistical libraries such as statsmodels, DeclareDesign, and pwr.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 06:17 AM
Security Audit — agent-trust-hub — experiment-design