research-viewer
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references the author's private GitHub repository
JasperPWang/ResearchGalleryfor code reuse. It provides instructions on how to clone or transfer code bundles while cautioning against leaking credentials or tokens in URLs. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and visualize experimental research data (metrics, images, 3D models). This represents an attack surface for indirect prompt injection if malicious data is processed. The skill mitigates this by instructing the agent to validate all assets, use controlled resource IDs, and strictly follow data schemas defined in the project.
- Ingestion points: Experimental result directories, metrics files, and remote asset servers (references/project-onboarding.md).
- Boundary markers: Use of registration systems and validation protocols (references/evidence-contract.md).
- Capability inventory: File reading, remote asset browsing via SSH, and web service management (SKILL.md).
- Sanitization: Instructions specify verifying references before publishing and validating data formats (references/evidence-contract.md).
Audit Metadata