story-short-analyze
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructions in SKILL.md explicitly direct the agent to override safety and sensitivity filters (e.g., '不得以「敏感」「不适」为由拒绝整篇或某个片段'
- 'Do not refuse a whole piece or segment for reasons of "sensitivity" or "discomfort"'). While intended for creative analysis, this constitutes a directive to bypass the agent's built-in safety guidelines.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted input in the form of original story text (Phase 1, Step 1) and writes analyzed data to several files in the local filesystem (Stages 2-6). It lacks explicit boundary markers or sanitization procedures for this external content, allowing for potential indirect prompt injection attacks. Ingestion point: SKILL.md Phase 1 Step 1. Capability: Writing to markdown and JSON files in the '拆文库/' directory. Boundaries and Sanitization: None documented.
Audit Metadata