project-visual-context
Pass
Audited by Gen Agent Trust Hub on Mar 17, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill ingests data from local JSON files and user-uploaded photos, which introduces a potential surface for indirect prompt injection.
- Ingestion points: Reads from project-config.json, plans-spatial.json, specs-quality.json, schedule.json, and directory.json.
- Boundary markers: No specific delimiters are used to isolate ingested data from agent instructions.
- Capability inventory: Read operations on project files and write operations to the visual-context.json file.
- Sanitization: Content is used directly without explicit sanitization or filtering.
- [DATA_EXFILTRATION]: The skill reads directory.json, which contains project contact details (architect and owner). This is used to understand design preferences and is considered normal functionality for this skill's use case.
Audit Metadata