kuroco-app-builder
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The file references/create-site.md contains curl commands that interact with the Kuroco platform's site registration API (https://rcms.g.kuroco.app). These commands facilitate new site setup and target the vendor's official infrastructure.
- [INDIRECT_PROMPT_INJECTION]: The skill maintains an attack surface for indirect prompt injection by processing external data and user requirements to generate application code and configurations.
- Ingestion points: Requirements collected from user input (SKILL.md 0-1), external mock response data (references/mock-contract.md), and platform-generated OpenAPI schemas (SKILL.md 3-4).
- Boundary markers: The orchestration uses a strict multi-phase checklist and a mock-first development pattern that separates unvalidated test data from actual API contracts.
- Capability inventory: Includes shell execution via curl for site registration (references/create-site.md) and extensive use of Admin MCP tools for backend configuration and content deployment (SKILL.md Phases 2 through 4).
- Sanitization: The skill recommends using a client-side mapping layer to transform raw external data into validated internal application types, as detailed in references/mock-contract.md.
Audit Metadata