agentic-ecology-ui
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The HTML template fetches font and icon assets from Google's public font services (fonts.googleapis.com and fonts.gstatic.com).
- [INDIRECT_PROMPT_INJECTION]: The skill's UI components ingest and render data from external vector databases, such as item titles and annotation labels, which represents a potential surface for indirect prompt injection. 1. Ingestion points: Data is loaded from JSON files via the /api/search and /api/labels endpoints in references/server.py and rendered by references/index.html. 2. Boundary markers: No explicit instructions are provided to the agent to treat data from the database as untrusted content. 3. Capability inventory: The associated server script has capabilities for file system read/write and network communication. 4. Sanitization: While the server uses os.path.basename to prevent path traversal, the content of the data fields is rendered using innerHTML in the results grid and label list within references/index.html without explicit sanitization.
Audit Metadata