design-study
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is focused on methodological rigor in medical research and contains no malicious code or exfiltration vectors.
- [COMMAND_EXECUTION]: The skill uses the Bash tool to run local scripts (adjustment_set_helper.py and verify.sh) for causal analysis and internal testing. These operations are limited to the skill's local environment and do not involve untrusted remote sources.
- [INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection by processing external documents. Evidence Chain: (1) Ingestion points: Phase 1 workflow extracts data from protocols, drafts, and notes. (2) Boundary markers: The workflow includes a 'Gate' for human-in-the-loop confirmation of the study summary. (3) Capability inventory: The skill has access to Bash, Write, and Edit tools. (4) Sanitization: No explicit sanitization of input text is implemented in the provided scripts. Given the human-in-the-loop requirement, this is classified as safe.
- [DYNAMIC_EXECUTION]: The included bash script uses heredocs to execute Python one-liners for JSON validation during local testing. This is a standard development practice for verification and does not pose a security risk as it uses known, local templates.
Audit Metadata