team-uidesign

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted content from the local codebase (e.g., package.json, ARCHITECTURE.md) and external skill outputs (e.g., ui-ux-pro-max), which are then interpolated into prompts for agents with write and execute capabilities.\n
  • Ingestion points: Found in roles/researcher/role.md (scanning project files) and SKILL.md (user arguments).\n
  • Boundary markers: None are present in the agent spawn templates to isolate interpolated data.\n
  • Capability inventory: The skill utilizes Write, Edit, Bash, and Agent spawning tools.\n
  • Sanitization: No evidence of validation or sanitization for ingested data before prompt construction.\n- [EXTERNAL_DOWNLOADS]: The skill relies on external intelligence from the ui-ux-pro-max skill and wiki-based search via the maestro CLI to inform its design and implementation decisions.\n- [COMMAND_EXECUTION]: The skill uses the Bash tool for environment detection and project stack identification, and this capability is enabled for all workers in the design pipeline.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 06:13 AM
Security Audit — agent-trust-hub — team-uidesign