Optimize
Orchestrates evo's parallel autoresearch loop to hill-climb a codebase benchmark
Test report
- Verdict
- Works with setup
- Score
- Tested
- Jul 20, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 1e57bc8
Fetched the repo tree via gh api and read the full 591-line canonical SKILL.md plus README; verified all 7 referenced files (sizing-the-round.md, evo-wait.md, glue.md, provider-matrix.md, cli-quick-reference.md, agents/ideator.md, workflows/evo-optimize.js) exist in-tree and found no exfiltration/base64/secret payloads on grep. Frontmatter is valid with name+description and uses ${CLAUDE_PLUGIN_ROOT}/relative paths (no hardcoded machine paths), but the skill is inert without the external evo-hq-cli plus an initialized workspace and committed baseline, so I docked install to 4. OUTPUT is NOT measured (not a measured loss): every step is an `evo` CLI call requiring an installed CLI + benchmark + likely GPU that I cannot provision, so output is set to unmeasured-neutral 5. Trigger 5/5 — the description is tightly scoped to an evo workspace with an explicit read-only exception, so all five test phrasings resolved correctly.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 4/5
- Triggers reliably 5/5
- Output vs. baseline 5/10
- Docs & honesty 4/5
What Optimize does
Drives the evo autoresearch optimization loop: after /evo:discover sets a benchmark and baseline, it plans rounds, spawns parallel semi-autonomous subagents in git worktrees to try code variants, verifies/gates results, and keeps what improves the score until it stalls. Triggers on /evo:optimize or plain-language "try some ideas / variants / use the GPUs / improve the current best" inside an initialized evo workspace. It is a loop-driver over the external evo CLI, not a standalone prompt.
How to install Optimize
uv tool install evo-hq-cli # then install the evo plugin/host hooks and run /evo:discover before /evo:optimize
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Optimize
-
/optimizeOrchestrates evo's parallel autoresearch loop to hill-climb a codebase benchmark
It also activates on plain-language prompts like these:
-
Try new variants to improve our current best result -
Run evo:optimize using the available GPUs -
Continue the evo search for a better candidate
Frequently asked questions
- Is the Optimize skill free?
- Yes. The skill itself is free from evo-hq/evo. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Optimize work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 20, 2026. Verdict: Works with setup. Fetched the repo tree via gh api and read the full 591-line canonical SKILL.md plus README; verified all 7 referenced files (sizing-the-round.md, evo-wait.md, glue.md, provider-matrix.md, cli-quick-reference.md, agents/ideator.md, workflows/evo-optimize.js) exist in-tree and found no exfiltration/base64/secret payloads on grep. Frontmatter is valid with name+description and uses ${CLAUDE_PLUGIN_ROOT}/relative paths (no hardcoded machine paths), but the skill is inert without the external evo-hq-cli plus an initialized workspace and committed baseline, so I docked install to 4. OUTPUT is NOT measured (not a measured loss): every step is an `evo` CLI call requiring an installed CLI + benchmark + likely GPU that I cannot provision, so output is set to unmeasured-neutral 5. Trigger 5/5 — the description is tightly scoped to an evo workspace with an explicit read-only exception, so all five test phrasings resolved correctly.
- What is the Optimize SkillProof Score?
- 7.2/10 — installs cleanly 4/5, triggers reliably 5/5, output vs. baseline 5/10, docs & honesty 4/5.
- How do I install Optimize?
- 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 Optimize 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.