Is the Auto Itera skill free?

Yes. The skill itself is free from clfhaha1234/auto-itera. SkillProof publishes the install command and an independent test verdict at no cost.

What is the Auto Itera SkillProof Score?

9.6/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 9/10, docs & honesty 5/5.

How do I install Auto Itera?

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 Auto Itera 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.

Auto Itera

Runs a disciplined baseline-vs-arms experiment to a sealed-test-set ship-or-kill verdict

by clfhaha1234 · clfhaha1234/auto-itera

Tested · Works

Install — copy the command Download source ↗

Test report

Verdict: Tested · Works
Score: 9.6/10
Tested: Jul 16, 2026
Environment: Claude Code 2.x (agent harness)
Upstream re-checked: Aug 10, 2026 · 4b58e67

Ran a real toy experiment (24 synthetic receipt rows, two vendor-extraction heuristics) two ways: the naive baseline scored both arms on all 24 rows and picked a winner from that single contaminated number, exactly the p-hacking anti-pattern the skill calls out; the skill-guided run enforced a real train/dev/sealed-test split, a Phase-3 pilot, per-slice diagnosis, and a genuine held-out pass - then rendered real PNGs with the bundled scripts/chart.py (matplotlib, executed for real, all three chart types produced valid images). The tiny N=5 test set landed the effect exactly at the pre-registered threshold, which is itself a live demonstration of the skill's own sizing-heuristic warning.

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 9/10
Docs & honesty 5/5

What Auto Itera does

Given a goal, candidate arms (prompts/models/architectures), and a pre-registered threshold, autonomously sources data, splits train/dev/held-out-test, scores arms in parallel, sprint-iterates behind a generalization gate, and writes a one-page conclusion doc with three real matplotlib charts. Triggers on 'should we use X or Y', 'compare these prompts/models', 'is this design better' style questions where a defensible verdict is needed, not on straightforward first-principles bug fixes.

How to install Auto Itera

git clone https://github.com/clfhaha1234/auto-itera
mkdir -p ~/.claude/skills
cp -r auto-itera/{SKILL.md,scripts,templates,examples} ~/.claude/skills/auto-itera/

Skills live in ~/.claude/skills/ (global) or .claude/skills/ (per-project). Restart Claude Code after installing.

Commands — how to trigger Auto Itera

/auto-itera Runs a disciplined baseline-vs-arms experiment to a sealed-test-set ship-or-kill verdict

It also activates on plain-language prompts like these:

Should we use prompt A or prompt B for extraction? Give me real evidence.
Compare these two model architectures with a proper held-out test set.
Is this new heuristic actually better, or am I just p-hacking myself?

Frequently asked questions

Is the Auto Itera skill free?: Yes. The skill itself is free from clfhaha1234/auto-itera. SkillProof publishes the install command and an independent test verdict at no cost.
Does Auto Itera work with Claude Code?: We tested it with Claude Code 2.x (agent harness) on Jul 16, 2026. Verdict: Tested · Works. Ran a real toy experiment (24 synthetic receipt rows, two vendor-extraction heuristics) two ways: the naive baseline scored both arms on all 24 rows and picked a winner from that single contaminated number, exactly the p-hacking anti-pattern the skill calls out; the skill-guided run enforced a real train/dev/sealed-test split, a Phase-3 pilot, per-slice diagnosis, and a genuine held-out pass - then rendered real PNGs with the bundled scripts/chart.py (matplotlib, executed for real, all three chart types produced valid images). The tiny N=5 test set landed the effect exactly at the pre-registered threshold, which is itself a live demonstration of the skill's own sizing-heuristic warning.
What is the Auto Itera SkillProof Score?: 9.6/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 9/10, docs & honesty 5/5.
How do I install Auto Itera?: 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 Auto Itera 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.