optim-agent Skill

Install or run

pip install -e ".[examples]"
pytest                     # runs tests/test_optim_agent.py

Client configuration

{
  "schema_version": 1,
  "suites": [
    {
      "id": "mnist",
      "result_glob": "docs/assets/mnist_curves_*_s*.json",
      "trials": 10,
      "seeds": [0, 1, 2, 3, 4],
      "metric": "sum of incumbent-best test error",
      "context_policy": "matched context and no-context agent variants"
    },
    {
      "id": "cifar10",
      "result_glob": "docs/assets/cifar10_curves_*_s*.json",
      "trials": 10,
      "seeds": [0, 1, 2, 3, 4],
      "metric": "sum of incumbent-best test error",
      "context_policy": "matched context and no-context agent variants"
    },
    {
      "id": "hard-functions",
      "result_glob": "docs/assets/hard_curves_*_s*.json",
      "trials": 10,
      "seeds": [0, 1, 2, 3, 4],
      "metric": "final incumbent objective value",
      "context_policy": "no supplied task context"
    },
    {
      "id": "rl-control",
      "result_glob": "docs/assets/rl_control_*_s*.json",
      "trials": 20,
      "seeds": [0, 1, 2, 3, 4],
      "environments": ["Acrobot-v1", "LunarLander-v3"],
      "model": "gpt-5.5",
      "agent_effort": "high",
      "history": 5,
      "selected_explicit_reasoning": false,
      "selected_qualitative_notes": false,
      "metric": "mean evaluation return; higher is better",
      "context_policy": "matched context and no-context agent variants"
    },
    {
      "id": "credit-default-benchmark",
      "result_glob": "docs/assets/credit_default_*_s*.json",
      "trials": 20,
      "seeds": [0, 1, 2, 3, 4],
      "dataset_id": 350,
      "dataset_sha256": "56c885f84457f6680f8438f02bfcdac9579323d8a94465ee5f26e32baa727602",
      "model": "gpt-5.5",
      "agent_effort": "high",
      "history": 20,
      "explicit_reasoning": true,
      "qualitative_notes": true,
      "metric": "validation incumbent log los

Compatible clients