無料GitHub で公開

optim-agent

Use when the user wants to optimize configurable system parameters against a measurable scalar objective, especially for model training, inference, quantitative strategies, reinforcement learning, scientific workflows, or other expensive black-box evaluations where reading the project can improve trial selection.

Optim-Agent/optim-agent

インストール方法を見る

含まれるファイル（200）

SKILL.md5.0 KB
.agents/plugins/marketplace.json374 B
.claude-plugin/marketplace.json438 B
.github/dependabot.yml269 B
.github/ISSUE_TEMPLATE/bug_report.md279 B
.github/ISSUE_TEMPLATE/feature_request.md193 B
.github/pull_request_template.md199 B
.github/workflows/ci.yml1.2 KB
.github/workflows/docs.yml710 B
.gitignore746 B
.pre-commit-config.yaml262 B
benchmarks/manifest.json2.1 KB
benchmarks/README.md3.7 KB
CHANGELOG.md1.7 KB
CITATION.cff630 B
CODE_OF_CONDUCT.md1013 B
CONTRIBUTING.md1.6 KB
docs/assets/cifar10_curves_GPT-5.5-medium_s0.json9.9 KB
docs/assets/cifar10_curves_GPT-5.5-medium_s1.json9.8 KB
docs/assets/cifar10_curves_GPT-5.5-medium_s2.json9.8 KB
docs/assets/cifar10_curves_GPT-5.5-medium_s3.json9.9 KB
docs/assets/cifar10_curves_GPT-5.5-medium_s4.json9.8 KB
docs/assets/cifar10_curves_GPT-5.5-medium-no-context_s0.json10.5 KB
docs/assets/cifar10_curves_GPT-5.5-medium-no-context_s1.json10.5 KB
docs/assets/cifar10_curves_GPT-5.5-medium-no-context_s2.json10.4 KB
docs/assets/cifar10_curves_GPT-5.5-medium-no-context_s3.json10.5 KB
docs/assets/cifar10_curves_GPT-5.5-medium-no-context_s4.json10.7 KB
docs/assets/cifar10_curves_Random_s0.json10.8 KB
docs/assets/cifar10_curves_Random_s1.json10.8 KB
docs/assets/cifar10_curves_Random_s2.json10.8 KB
docs/assets/cifar10_curves_Random_s3.json10.7 KB
docs/assets/cifar10_curves_Random_s4.json10.9 KB
docs/assets/cifar10_curves_TPE_s0.json10.7 KB
docs/assets/cifar10_curves_TPE_s1.json10.8 KB
docs/assets/cifar10_curves_TPE_s2.json10.8 KB
docs/assets/cifar10_curves_TPE_s3.json10.7 KB
docs/assets/cifar10_curves_TPE_s4.json10.8 KB
docs/assets/classification_benchmarks.png102.4 KB
docs/assets/credit_card.png93.2 KB
docs/assets/credit_default_GP-BO_s0.json6.9 KB
docs/assets/credit_default_GP-BO_s1.json7.0 KB
docs/assets/credit_default_GP-BO_s2.json6.9 KB
docs/assets/credit_default_GP-BO_s3.json6.9 KB
docs/assets/credit_default_GP-BO_s4.json6.9 KB
docs/assets/credit_default_GPT-5.5_s0.json6.5 KB
docs/assets/credit_default_GPT-5.5_s1.json6.6 KB
docs/assets/credit_default_GPT-5.5_s2.json6.6 KB
docs/assets/credit_default_GPT-5.5_s3.json6.5 KB
docs/assets/credit_default_GPT-5.5_s4.json6.6 KB
docs/assets/credit_default_GPT-5.5-no-context_s0.json6.0 KB
docs/assets/credit_default_GPT-5.5-no-context_s1.json6.1 KB
docs/assets/credit_default_GPT-5.5-no-context_s2.json6.1 KB
docs/assets/credit_default_GPT-5.5-no-context_s3.json5.9 KB
docs/assets/credit_default_GPT-5.5-no-context_s4.json6.0 KB
docs/assets/credit_default_Random_s0.json6.7 KB
docs/assets/credit_default_Random_s1.json6.7 KB
docs/assets/credit_default_Random_s2.json6.7 KB
docs/assets/credit_default_Random_s3.json6.7 KB
docs/assets/credit_default_Random_s4.json6.7 KB
docs/assets/credit_default_TPE_s0.json6.7 KB
docs/assets/credit_default_TPE_s1.json6.7 KB
docs/assets/credit_default_TPE_s2.json6.7 KB
docs/assets/credit_default_TPE_s3.json6.7 KB
docs/assets/credit_default_TPE_s4.json6.7 KB
docs/assets/hard_benchmarks_free.png127.0 KB
docs/assets/hard_benchmarks_tier.png148.7 KB
docs/assets/hard_curves_Big-pickle_s0.json2.4 KB
docs/assets/hard_curves_Big-pickle_s1.json2.4 KB
docs/assets/hard_curves_Big-pickle_s2.json2.4 KB
docs/assets/hard_curves_Big-pickle_s3.json2.4 KB
docs/assets/hard_curves_Big-pickle_s4.json2.4 KB
docs/assets/hard_curves_DeepSeek-V4-Flash_s0.json2.4 KB
docs/assets/hard_curves_DeepSeek-V4-Flash_s1.json2.4 KB
docs/assets/hard_curves_DeepSeek-V4-Flash_s2.json2.5 KB
docs/assets/hard_curves_DeepSeek-V4-Flash_s3.json2.4 KB
docs/assets/hard_curves_DeepSeek-V4-Flash_s4.json2.5 KB
docs/assets/hard_curves_GLM-5.2_s0.json2.4 KB
docs/assets/hard_curves_GLM-5.2_s1.json2.4 KB
docs/assets/hard_curves_GLM-5.2_s2.json2.4 KB
docs/assets/hard_curves_GLM-5.2_s3.json2.5 KB
docs/assets/hard_curves_GLM-5.2_s4.json2.4 KB
docs/assets/hard_curves_GP-BO_s0.json2.8 KB
docs/assets/hard_curves_GP-BO_s1.json2.8 KB
docs/assets/hard_curves_GP-BO_s2.json2.8 KB
docs/assets/hard_curves_GP-BO_s3.json2.8 KB
docs/assets/hard_curves_GP-BO_s4.json2.8 KB
docs/assets/hard_curves_GPT-5.5_s0.json2.5 KB
docs/assets/hard_curves_GPT-5.5_s1.json2.5 KB
docs/assets/hard_curves_GPT-5.5_s2.json2.4 KB
docs/assets/hard_curves_GPT-5.5_s3.json2.5 KB
docs/assets/hard_curves_GPT-5.5_s4.json2.5 KB
docs/assets/hard_curves_Kimi-K3_s0.json2.4 KB
docs/assets/hard_curves_Kimi-K3_s1.json2.5 KB
docs/assets/hard_curves_Kimi-K3_s2.json2.4 KB
docs/assets/hard_curves_Kimi-K3_s3.json2.5 KB
docs/assets/hard_curves_Kimi-K3_s4.json2.5 KB
docs/assets/hard_curves_MiMo-v2.5_s0.json2.4 KB
docs/assets/hard_curves_MiMo-v2.5_s1.json2.4 KB
docs/assets/hard_curves_MiMo-v2.5_s2.json2.4 KB
docs/assets/hard_curves_MiMo-v2.5_s3.json2.4 KB
docs/assets/hard_curves_MiMo-v2.5_s4.json2.5 KB
docs/assets/hard_curves_Minimax-M3_s0.json2.4 KB
docs/assets/hard_curves_Minimax-M3_s1.json2.4 KB
docs/assets/hard_curves_Minimax-M3_s2.json2.4 KB
docs/assets/hard_curves_Minimax-M3_s3.json2.4 KB
docs/assets/hard_curves_Minimax-M3_s4.json2.4 KB
docs/assets/hard_curves_Nemotron-3-Ultra_s0.json2.4 KB
docs/assets/hard_curves_Nemotron-3-Ultra_s1.json2.4 KB
docs/assets/hard_curves_Nemotron-3-Ultra_s2.json2.5 KB
docs/assets/hard_curves_Nemotron-3-Ultra_s3.json2.5 KB
docs/assets/hard_curves_Nemotron-3-Ultra_s4.json2.5 KB
docs/assets/hard_curves_Opus-4.8_s0.json2.5 KB
docs/assets/hard_curves_Opus-4.8_s1.json2.5 KB
docs/assets/hard_curves_Opus-4.8_s2.json2.5 KB
docs/assets/hard_curves_Opus-4.8_s3.json2.5 KB
docs/assets/hard_curves_Opus-4.8_s4.json2.4 KB
docs/assets/hard_curves_Random_s0.json3.1 KB
docs/assets/hard_curves_Random_s1.json3.1 KB
docs/assets/hard_curves_Random_s2.json3.0 KB
docs/assets/hard_curves_Random_s3.json3.0 KB
docs/assets/hard_curves_Random_s4.json3.1 KB
docs/assets/hard_curves_Sonnet-5_s0.json2.4 KB
docs/assets/hard_curves_Sonnet-5_s1.json2.4 KB
docs/assets/hard_curves_Sonnet-5_s2.json2.4 KB
docs/assets/hard_curves_Sonnet-5_s3.json2.4 KB
docs/assets/hard_curves_Sonnet-5_s4.json2.4 KB
docs/assets/hard_curves_TPE_s0.json3.1 KB
docs/assets/hard_curves_TPE_s1.json3.0 KB
docs/assets/hard_curves_TPE_s2.json3.1 KB
docs/assets/hard_curves_TPE_s3.json3.1 KB
docs/assets/hard_curves_TPE_s4.json3.0 KB
docs/assets/lunarlander_landing.json1.4 KB
docs/assets/lunarlander_policy.gif462.7 KB
docs/assets/mnist_curves_GPT-5.5-medium_s0.json10.0 KB
docs/assets/mnist_curves_GPT-5.5-medium_s1.json10.0 KB
docs/assets/mnist_curves_GPT-5.5-medium_s2.json10.0 KB
docs/assets/mnist_curves_GPT-5.5-medium_s3.json10.0 KB
docs/assets/mnist_curves_GPT-5.5-medium_s4.json10.0 KB
docs/assets/mnist_curves_GPT-5.5-medium-no-context_s0.json10.8 KB
docs/assets/mnist_curves_GPT-5.5-medium-no-context_s1.json10.8 KB
docs/assets/mnist_curves_GPT-5.5-medium-no-context_s2.json10.7 KB
docs/assets/mnist_curves_GPT-5.5-medium-no-context_s3.json10.7 KB
docs/assets/mnist_curves_GPT-5.5-medium-no-context_s4.json10.8 KB
docs/assets/mnist_curves_Random_s0.json11.0 KB
docs/assets/mnist_curves_Random_s1.json11.1 KB
docs/assets/mnist_curves_Random_s2.json11.1 KB
docs/assets/mnist_curves_Random_s3.json11.0 KB
docs/assets/mnist_curves_Random_s4.json11.0 KB
docs/assets/mnist_curves_TPE_s0.json11.0 KB
docs/assets/mnist_curves_TPE_s1.json11.0 KB
docs/assets/mnist_curves_TPE_s2.json10.9 KB
docs/assets/mnist_curves_TPE_s3.json11.1 KB
docs/assets/mnist_curves_TPE_s4.json11.0 KB
docs/assets/optim-agent-logo-dark.svg4.5 KB
docs/assets/optim-agent-logo-light.svg4.5 KB
docs/assets/optim-agent-mark.svg633 B
docs/assets/optim-agent-overview.png90.6 KB
docs/assets/optimization_trajectory.gif493.7 KB
docs/assets/rl_control_GPT-5.5_s0.json6.9 KB
docs/assets/rl_control_GPT-5.5_s1.json6.9 KB
docs/assets/rl_control_GPT-5.5_s2.json6.9 KB
docs/assets/rl_control_GPT-5.5_s3.json6.9 KB
docs/assets/rl_control_GPT-5.5_s4.json7.0 KB
docs/assets/rl_control_GPT-5.5-no-context_s0.json4.9 KB
docs/assets/rl_control_GPT-5.5-no-context_s1.json5.0 KB
docs/assets/rl_control_GPT-5.5-no-context_s2.json4.8 KB
docs/assets/rl_control_GPT-5.5-no-context_s3.json4.9 KB
docs/assets/rl_control_GPT-5.5-no-context_s4.json4.9 KB
docs/assets/rl_control_Random_s0.json5.7 KB
docs/assets/rl_control_Random_s1.json5.7 KB
docs/assets/rl_control_Random_s2.json5.7 KB
docs/assets/rl_control_Random_s3.json5.8 KB
docs/assets/rl_control_Random_s4.json5.7 KB
docs/assets/rl_control_TPE_s0.json5.7 KB
docs/assets/rl_control_TPE_s1.json5.7 KB
docs/assets/rl_control_TPE_s2.json5.7 KB
docs/assets/rl_control_TPE_s3.json5.8 KB
docs/assets/rl_control_TPE_s4.json5.7 KB
docs/assets/rl_control.png77.8 KB
docs/assets/social-preview.png93.6 KB
docs/i18n/README_DE.md22.3 KB
docs/i18n/README_ES.md22.4 KB
docs/i18n/README_FR.md22.9 KB
docs/i18n/README_JA.md23.6 KB
docs/i18n/README_KO.md22.5 KB
docs/i18n/README_PT.md22.2 KB
docs/i18n/README_RU.md27.1 KB
docs/i18n/README_ZH.md20.8 KB
docs/index.html37.1 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging-retry/ABORTED_RUN_MANIFEST.40b13e0d.json4.9 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging-retry/EXECUTION_MANIFEST.json5.1 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging-retry/WORKER_BLOCKED.TASK-001.json520 B
docs/optim-plans/2026-07-27-summary-agent-and-table-logging-retry/WORKER_BLOCKED.TASK-001.reporting.py4.2 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging-retry2/EXECUTION_MANIFEST.json5.6 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging-retry3/EXECUTION_MANIFEST.json6.3 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging-retry4/EXECUTION_MANIFEST.json5.9 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging/PLAN_v1_reviewer_comments.md3.0 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging/PLAN_v1.md9.6 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging/PLAN_v2_reviewer_comments.md2.8 KB
docs/optim-plans/2026-07-27-summary-agent-and-table-logging/PLAN_v2.md12.4 KB

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

optim-agent

Act as the sampler inside any coding-agent session: Claude Code, Codex, OpenCode/OpenClaw, or another agent that can read project files and run shell commands. Read the project to understand parameter meaning and interactions, propose one configuration, run the real evaluator, and record the result through optim-agent's ask/tell API. Let the measured objective, not the agent's intuition, decide what works.

Load the workflow

Use this file as the operating guide for the active coding agent. In Codex, it can be installed directly from GitHub:

$skill-installer install https://github.com/Optim-Agent/optim-agent

In Claude Code, OpenCode/OpenClaw, or another coding-agent environment, place this repository or SKILL.md in the agent-visible workspace and ask the agent to follow the optim-agent workflow. The workflow does not depend on Codex-only APIs; it needs file access, shell access, and Python.

Ensure the Python package is importable. Choose one source; do not install both:

# Stable release from PyPI
python -m pip install optim-agent

# Latest source from GitHub
python -m pip install "optim-agent @ git+https://github.com/Optim-Agent/optim-agent.git"

For a reproducible GitHub install, append @<tag-or-commit> after .git.

Workflow

Understand the system. Read the evaluation entry point and every file that defines the target parameters. Record each parameter's type, legal range, semantics, interactions, and operational constraints.
Define the experiment. Confirm the scalar objective, minimize or maximize, trial budget, evaluation command, runtime/cost limit, and fixed workload or seed. For multiple metrics or hard constraints, agree on one scalar feasibility or penalty rule before running trials.
Establish a baseline. Evaluate the current/default configuration with the same command and environment used for every later trial.

Initialize or resume. Keep artifacts in the repository's ignored .optim-agent-runs/ directory:

if git rev-parse --git-dir >/dev/null 2>&1 && ! git check-ignore -q .optim-agent-runs/; then
  printf '/.optim-agent-runs/\n' >> "$(git rev-parse --git-path info/exclude)"
fi

from pathlib import Path
import optim_agent as oa

run_dir = Path(".optim-agent-runs")
run_dir.mkdir(exist_ok=True)
study = oa.create_study(
    direction="minimize",
    storage=run_dir / "skill-study.json",
    seed=0,
)
print([(t.params, t.value, t.state) for t in study.trials])

Run one informed trial. Choose parameters from code understanding and all completed history, then use explicit ask/tell:
```
params = {"threshold": 0.72, "budget": 80}
trial = study.ask(params)
try:
    value = evaluate_system(**trial.params)
except Exception:
    study.tell(trial, state="failed")
    raise
else:
    study.tell(trial, value)
```
For a deliberately stopped trial, report the latest valid intermediate metric first, then call study.tell(trial, state="pruned").
Select the next point. Avoid accidental repeats, explore broadly before exploiting, respect bounds and constraints, and treat failed regions as evidence. If the evaluator is noisy, repeat promising configurations under the same workload before declaring a winner.
Stop and report. Stop at the approved budget or stopping condition. Report the baseline, best value and parameters, trial count, failed/pruned trials, convergence trend, and exact reproduction command.

Recovery

JSON storage records a trial when study.tell runs. Before launching an expensive external evaluation, save its parameters, command, and output path in a per-trial directory under .optim-agent-runs/. After interruption, inspect that output before rerunning: if a valid result exists, recreate the same point with study.ask(params) and record it; otherwise rerun it deliberately.

Use SQLite storage (skill-study.db) only when the user explicitly wants multiple processes. Sequential trials are the default because each proposal should use the complete prior history.

Rules

Use ask/tell in skill mode; do not delegate proposal selection to AgentSampler when the session agent is meant to read and reason over code.
Keep evaluation inputs and outputs isolated from production configuration.
Never fabricate, infer, or manually improve an objective value.
Record crashes as failed; record intentional early stops as pruned.
Preserve the study and trial artifacts so the result is auditable and resumable.
Do not tune secrets, credentials, or unbounded parameters.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

ログインしてレビューを書く

同じリポジトリのスキル

概要と使いどころ

optim-agent

無料

Use when optimizing configurable system parameters against a measurable scalar objective.

日本語の概要は準備中です。原文の説明を表示しています。

Optim-Agent/optim-agent8012026年8月15日更新

Optim-Agent のスキルをすべて見る

このスキルの問題を報告する

インストール

コマンドで入れる

npx skills add https://github.com/Optim-Agent/optim-agent --skill optim-agent

ターミナルに貼り付けて実行します（Node.js が必要です）。

Claude に頼む

次の Agent Skill をインストールしてください。
https://github.com/Optim-Agent/optim-agent

まず SKILL.md と付属ファイルの中身を確認し、不審な処理（外部への送信、認証情報の読み取りなど）がないことを確かめてから、~/.claude/skills/optim-agent/ に配置してください。

この文章をそのまま Claude Code に送ると、中身を確認してから入れてくれます。

GitHub でソースを見る

リポジトリ: Optim-Agent/optim-agent
スター: 801
フォーク: 43
カテゴリ: リサーチ・学習
ライセンス: MIT
本文の言語: 英語
最終更新: 2026年8月15日

スキルはエージェントへの指示とスクリプトの集まりです。cccskills は内容の安全性を保証していません。入れる前に中身を確認してください。