yao-cold-ledger-skill
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze user-provided text, creating a surface for potential indirect prompt injection attacks. Ingestion points: Reference texts (参考文本) and user-provided story materials (素材) as specified in SKILL.md. Boundary markers: The skill contains explicit logic instructing the agent to treat document-embedded text solely as content for analysis ('待分析内容') and requires direct user dialogue to change tasks. Capability inventory: Limited to text analysis and story generation; the skill utilizes no network, file-writing, or persistent storage capabilities. Sanitization: Employs logical separation between the narrative persona and external input to prevent accidental obedience to embedded instructions.
- [SAFE]: A thorough review of all skill components, including the methodology, sentence patterns, and fictional examples, confirms that the skill performs only benign text-processing tasks. No hidden obfuscation, remote code execution, or credential harvesting patterns were found.
Audit Metadata