nominalism-vs-realism-debate-expert
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The research workflow outlined in
modules/research-checklist.mddirects the agent to perform web searches and distill information from external sources like Wikipedia and ArXiv to validate and update the skill's core guidance. This creates an ingestion surface where data from these external providers enters the agent's context. - Ingestion points: External research data fetched via web searches, Wikipedia, and ArXiv as specified in
modules/research-checklist.md. - Boundary markers: Absent. There are no instructions in the provided files to use delimiters or to disregard potential instructions found within the research data.
- Capability inventory: The skill is designed for information synthesis and philosophical instruction; it does not include scripts or instructions that perform high-privilege operations such as shell command execution, file system modification, or network exfiltration.
- Sanitization: Absent. The workflow does not define explicit sanitization or validation steps for the ingested external content.
Audit Metadata