novel-game-analyze
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, such as novels and writing projects, to extract game design facts. This constitutes an ingestion point for potential indirect prompt injection attacks if the source material contains malicious instructions designed to override agent behavior.
- Ingestion points: Raw novels, analysis libraries, and writing projects provided as input in SKILL.md.
- Boundary markers: The instructions mandate the use of structured tables (Source Bible) and specific textual citations, which helps maintain a boundary between raw data and the agent's analysis logic.
- Capability inventory: The skill performs text analysis and file generation within the local workspace. No high-risk tools such as network access or shell execution are requested.
- Sanitization: The protocol requires 'Fact Fidelity' and mandates cross-referencing all extracted facts with original text positions, which acts as a verification step against adversarial content.
Audit Metadata