fusion-to-publish-v2

Pass

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Executes local diagnostic scripts (detect-project.sh and scan-components.sh) to analyze the project structure, detect installed SDK versions, and identify unregistered React components.
  • [EXTERNAL_DOWNLOADS]: Recommends the installation of official vendor packages including @builder.io/react, @builder.io/dev-tools, and @builder.io/sdk from the NPM registry to enable CMS functionality.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes untrusted data from the user's codebase:
  • Ingestion points: Reads local React component files (.tsx, .jsx) to extract property interfaces.
  • Boundary markers: No explicit delimiters or instructions are used to separate source code analysis from the generation logic.
  • Capability inventory: The agent is empowered to write new files (builder-registry.ts, catch-all routes) and modify existing build configurations (next.config.ts).
  • Sanitization: Extracted prop names and types are interpolated directly into code templates for the component registry without a dedicated validation or escaping layer.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 26, 2026, 03:04 AM
Security Audit — agent-trust-hub — fusion-to-publish-v2