tw-research-ethics-reviewer
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-supplied research parameters to generate structured ethics reports and consent forms, creating a potential surface for indirect prompt injection.
- Ingestion points: User-provided research design, data collection methods, and participant details collected in Step 1 and Step 2 are processed to generate outputs.
- Boundary markers: Absent; user input is directly interpolated into template placeholders (e.g., [研究題目], [研究目的]) in Step 4 and Step 6.
- Capability inventory: The skill environment is configured with access to Bash, Read, and Write tools as specified in the allowed-tools frontmatter.
- Sanitization: No explicit sanitization or instruction-filtering logic is present to prevent the agent from executing instructions embedded within the user-provided research data.
Audit Metadata