MCP Server for executing Python code in an isolated REPL namespace with captured stdout/stderr
Single tool with minimal schema and description. Tool description is 104 chars (within 10-1024 range) but generic. No output schema documented. No error handling guidance. No parameter constraints or validation hints. Tool name 'run_python_code' is verb-based but vague about isolation guarantees and safety model. Missing all composition patterns (no chaining IDs, no pagination, no structured output guidance).
Execute Python code in an isolated REPL namespace with captured stdout.
Parameter 'code' lacks description in schema. LLM cannot infer constraints, format expectations, or safety boundaries.
No output schema documented. Response is JSON-serialized dict with status/stdout/stderr/eval_result/error/elapsed_sec/active_variables. LLM cannot plan downstream actions or extract structured data reliably.
No error handling guidance. Tool can fail with SyntaxError, NameError, RuntimeError, etc. but response only includes generic 'status' and 'error' fields.
Irreversible operations (code execution with side effects) lack confirmation or dry-run pattern. Tool description does not warn about destructive capability or isolation limits.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
Tool description does not clarify what 'isolated namespace' means to the LLM. Does it prevent file I/O? Network access? Import restrictions? Unclear boundaries invite misuse.