redesign-existing-projects
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: Surface for Indirect Prompt Injection. The skill directs the agent to 'Scan — Read the codebase', which creates an entry point for untrusted data. If the project code contains hidden instructions, the agent could potentially execute them while performing 'targeted upgrades'.\n
- Ingestion points: Project source code files (SKILL.md).\n
- Boundary markers: The instructions do not define clear boundaries or isolation directives for content found within the code files.\n
- Capability inventory: The agent has capabilities to modify the filesystem to apply design fixes across the project.\n
- Sanitization: No specific validation or sanitization of the scanned project content is performed before diagnostic processing.\n- [EXTERNAL_DOWNLOADS]: Reference to a well-known placeholder service. The skill recommends using
https://picsum.photosto generate placeholder background images when real assets are unavailable.
Audit Metadata