Module eval_tool 

Eval tool — persistent-kernel code execution (scaffold). Eval tool — execute code and capture output.

Provides a multi-language code-execution environment (Python and JavaScript today). Each call writes code to a temp file, executes it via the appropriate runtime (python3 or bun/node), and captures stdout/stderr + exit code.

Each call runs in a fresh process — persistent kernel sessions across calls are a future enhancement. For interactive or multi-step sessions, use bash with python3 -i or bun -i.

Structs

EvalTool

eval agent tool — run code and capture output.