MCP server for evaluating model constitutional safety and harmlessness using Constitutional AI principles with benchmark scoring
Critical failures across all dimensions. The single tool 'execute_skill_action' has no visible input schema, no parameter descriptions, and a generic 1-line description (23 chars). The mcp_server.py file contains only a stub that prints JSON and exits, it does not implement the MCP protocol at all. No tool registration, no request/response handling, no error handling. The client.py and example_usage.py files are disconnected from the server and do not demonstrate MCP integration. This is not a functional MCP server.
Execute Constitutional AI harmlessness evaluation benchmark skill action
Tool description is 23 characters ('Execute Constitutional AI harmlessness evaluation benchmark skill action'). This is generic and does not explain WHAT the tool does, WHEN to use it, or what it returns.
mcp_server.py does not implement MCP protocol. It only prints a JSON status message and exits. No request/response loop, no tool invocation handling, no message parsing. This is a stub, not a server.
No parameter descriptions or types. The tool accepts parameters (inferred from client.py: model_identifier, constitution_principles_count) but the server has no schema defining them, their types, constraints, or descriptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 18 | 2026-07-28+ | v2 |
No error handling, no recovery guidance, no validation. The client.py returns hardcoded mock data; the server does not validate inputs or handle failures.