sumeru-outline
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided creative concepts and novel themes, representing an indirect prompt injection surface.
- Ingestion points: User-supplied novel ideas and prompts (e.g., '帮我写个小说大纲') are ingested into the agent context as described in SKILL.md.
- Boundary markers: The skill uses structured Markdown templates and explicit output segments to separate generated content, though it lacks strict delimiters for raw user input.
- Capability inventory: The skill writes structured JSON and Markdown data to the .sumeru/outline/ directory and coordinates parallel chapter generation via multiple sub-agents.
- Sanitization: The skill features a mandatory AI-powered 'Compliance Check' mechanism that identifies potential real-world names or locations and replaces them with fictional alternatives, significantly reducing the risk of malicious entity injection or legal infringement.
Audit Metadata