nocobase-ai-builder
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is instructed to ingest and follow instructions from project-specific files such as
AGENTS.md. This represents a potential vulnerability where malicious prompts within a repository could manipulate the agent's behavior. - Ingestion points: The agent processes project source files, Swagger API documentation, and user-provided requirements.
- Boundary markers: The skill instructions do not define strict isolation or sanitization for instructions encountered within the local repository source.
- Capability inventory: The skill has extensive filesystem permissions to modify code and execute NocoBase CLI (
nb) commands for synchronization and deployment. - Sanitization: There is no documented process for validating or filtering prompts ingested from project-level documentation.
- [COMMAND_EXECUTION]: The skill uses the NocoBase CLI (
nb) for primary tasks such as pulling source repositories (nb portal pull), discovery of API endpoints (nb api swagger), and database resource querying. These are core functionalities but establish a significant system interaction surface. - [DYNAMIC_EXECUTION]: The skill involves dynamic code interaction including merging template source updates, executing production builds, and running automated E2E test suites against modified code as part of its verification process.
Audit Metadata