paper-claim-audit
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill performs standard file-based auditing tasks and does not contain malicious code or exfiltration patterns.
- [SAFE]: The use of a fresh thread for the reviewer model is a positive security design that prevents session-to-session context leakage.
- [SAFE]: The skill exposes a surface for indirect prompt injection because it reads untrusted paper and result files and passes them to an LLM. However, the risk is low due to the specific audit protocol and structured output requirements. Evidence chain: 1. Ingestion: .tex and data files in paper/ and results/ directories; 2. Boundary markers: Absent; 3. Capabilities: Write, Bash, Agent; 4. Sanitization: Absent.
Audit Metadata