Optim-Agent
Disciplined ask/tell loop for black-box parameter tuning against a measurable objective
Test report
- Verdict
- Tested · Works
- Score
- Tested
- Jul 17, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 544e334
Ran a real 12-evaluation black-box minimisation both ways: the skill's disciplined ask/tell loop (baseline first, explore then exploit over recorded history) reached objective 0.03 right next to the true optimum, while unaided intuition-guessing stalled at 1.58 at the same budget -- about 98% lower. Worth flagging that the pip package is essentially an ask/tell ledger (its only samplers are Random and AgentSampler), so the win comes from the prescribed workflow, not a built-in Bayesian optimiser.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 7/10
- Docs & honesty 4/5
What Optim-Agent does
Turns the agent into a sequential sampler for expensive black-box optimization: read the code, propose one config, run the real evaluator, record via an ask/tell study, and let the measured objective decide. Use for hyperparameter, inference, quant-strategy, RL or scientific-workflow tuning. Backed by the pip package optim-agent.
How to install Optim-Agent
python -m pip install optim-agent
git clone https://github.com/Optim-Agent/optim-agent.git
mkdir -p ~/.claude/skills/optim-agent
cd optim-agent && cp SKILL.md ~/.claude/skills/optim-agent/SKILL.md
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Optim-Agent
-
/optim-agentDisciplined ask/tell loop for black-box parameter tuning against a measurable objective
It also activates on plain-language prompts like these:
-
Help me tune XGBoost hyperparameters to minimize validation loss -
Need to find the best backtest threshold and budget for our Sharpe ratio -
Tune the reward-shaping parameters in our RL agent against episode return
Frequently asked questions
- Is the Optim-Agent skill free?
- Yes. The skill itself is free from Optim-Agent/optim-agent. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Optim-Agent work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 17, 2026. Verdict: Tested · Works. Ran a real 12-evaluation black-box minimisation both ways: the skill's disciplined ask/tell loop (baseline first, explore then exploit over recorded history) reached objective 0.03 right next to the true optimum, while unaided intuition-guessing stalled at 1.58 at the same budget -- about 98% lower. Worth flagging that the pip package is essentially an ask/tell ledger (its only samplers are Random and AgentSampler), so the win comes from the prescribed workflow, not a built-in Bayesian optimiser.
- What is the Optim-Agent SkillProof Score?
- 8.4/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 7/10, docs & honesty 4/5.
- How do I install Optim-Agent?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Optim-Agent with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.