nocobase-create-app

Pass

Audited by Gen Agent Trust Hub on Oct 17, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches NocoBase application templates and packages from the official vendor registry at https://npm.nocobase.ai. This infrastructure is managed by the skill author.
  • [REMOTE_CODE_EXECUTION]: Uses pnpm create @nocobase/app to download and execute initialization scripts from the vendor's package registry.
  • [COMMAND_EXECUTION]: Executes several local commands for environment setup (node, pnpm), application configuration (pnpm nocobase config), and starting development processes (pnpm dev). It also includes troubleshooting commands like rm -rf node_modules.
  • [INDIRECT_PROMPT_INJECTION]: The skill is instructed to read AGENTS.md and follow instructions from within the newly created application directory. If the project source or template contains malicious instructions, they could influence the agent's behavior.
  • Ingestion points: Reads AGENTS.md and config.yml from the created project directory.
  • Boundary markers: None explicitly defined for the newly discovered AGENTS.md content.
  • Capability inventory: Subprocess execution via pnpm, local file system access, and network requests via pnpm and curl.
  • Sanitization: No explicit sanitization or validation of the contents of the newly read project files before processing instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 17, 2026, 12:19 PM