cocos-creator-mcp-server

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The documentation provides manual installation instructions involving cloning an external repository from https://github.com/DaxianLee/cocos-mcp-server.git.
  • [COMMAND_EXECUTION]: The skill exposes tools to build and run the current Cocos Creator project (project_manage, project_build_system), allowing the AI to initiate local process execution.
  • [DATA_EXFILTRATION]: The skill grants access to local development data including console logs and project-specific files. The debug_logs tool allows searching through log files, and the asset_manage tool allows importing files from local paths, which could be leveraged to access sensitive information if the assistant is manipulated.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a high attack surface for indirect injection as it ingests and processes various project-related data.
  • Ingestion points: Reads scene hierarchies, node names, component properties, and console log outputs.
  • Boundary markers: No explicit delimiters or instruction-ignore warnings are specified in the tool descriptions.
  • Capability inventory: The skill can write files (scenes/prefabs), execute commands (build/run), and perform network operations via the MCP server.
  • Sanitization: There is no evidence of sanitization or validation of the data ingested from the editor environment before it is processed by the AI.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 10:44 PM
Security Audit — agent-trust-hub — cocos-creator-mcp-server