paper-review-zh-cn
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of instructional text and review frameworks. It does not perform network operations, access sensitive local files, or execute external code. All references are to internal documentation files.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external academic manuscripts, which constitutes an attack surface for indirect prompt injection. 1. Ingestion points: Academic manuscripts and accompanying products specified in SKILL.md. 2. Boundary markers: The skill instructs the agent to anchor comments to specific text segments and explicitly warns against following embedded instructions or fabricating content in overlay-contract.md. 3. Capability inventory: No shell commands, network operations, or dynamic code execution capabilities are defined within the skill files. 4. Sanitization: The skill emphasizes evidence verification and checking for observational features of AI-generated text rather than assuming authorship.
Audit Metadata