Is the Codex Autoresearch skill free?

Yes. The skill itself is free from TheGreenCedar/codex-autoresearch. SkillProof publishes the install command and an independent test verdict at no cost.

What is the Codex Autoresearch SkillProof Score?

8.8/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 8/10, docs & honesty 4/5.

How do I install Codex Autoresearch?

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 Codex Autoresearch 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.

Codex Autoresearch

Measured, resumable optimization loop with baseline, correctness gate, and evidence ledger

by TheGreenCedar · TheGreenCedar/codex-autoresearch

Tested · Works

Install — copy the command Download source ↗

Test report

Verdict: Tested · Works
Score: 8.8/10
Tested: Jul 21, 2026
Environment: Claude Code 2.x (agent harness)
Upstream re-checked: Aug 10, 2026 · f0fac23

I built a toy Node project (sumTo via array+reduce, a bench.mjs printing METRIC seconds= and a checks.mjs), then ran the loop two ways under a temp HOME. Baseline (no skill): I just rewrote the function to an accumulator loop and re-ran bench (0.7297s -> 0.3652s) — a faster file, no record. Skill: setup wrote a session (autoresearch.md/.jsonl/.sh), doctor --check-benchmark confirmed the METRIC contract, next+log captured baseline 0.7234, then the keep packet logged 0.3542 with checks passing (exit 0, "checks OK"), created scoped git commit 8d4e192 touching only lib/sum.mjs, attached an ASI note (hypothesis/evidence/next-action), and finalize-preview produced a review-branch plan flagging maturity "experimental". Decisive difference: when I fed a faster-but-wrong change (closed-form +1), the packet's checks exited 1 and the CLI forced status checks_failed as the only allowed status, refusing the keep — the baseline path has no such regression gate.

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 8/10
Docs & honesty 4/5

What Codex Autoresearch does

Turns a vague "make it faster/smaller" request into a measured loop: it records a benchmark baseline, runs one scoped change per packet, re-runs correctness checks, and logs each result (metric, decision, evidence, next action) to a resumable session file with a scoped git commit. Triggers on benchmark-driven optimization (test runtime, bundle size, latency, memory), qualitative quality-gap research, resuming an existing autoresearch session, or preparing a review-branch finalization. Backed by a Node CLI (autoresearch.mjs) that runs the benchmark and checks commands locally.

How to install Codex Autoresearch

git clone --depth 1 https://github.com/TheGreenCedar/codex-autoresearch.git /tmp/codex-autoresearch-src
mkdir -p ~/.claude/skills
cp -R /tmp/codex-autoresearch-src/plugins/codex-autoresearch/skills/codex-autoresearch ~/.claude/skills/codex-autoresearch
# NOTE: this is a Codex PLUGIN, not a standalone skill. The SKILL.md drives a Node CLI that
# lives OUTSIDE the skill folder at /tmp/codex-autoresearch-src/plugins/codex-autoresearch/scripts/autoresearch.mjs
# Requirements: Node >= 24. On first run the CLI hydrates a runtime (downloads the matching
# GitHub release tarball with sha256 verification, or build locally from the clone).
# Run commands from the package root, e.g.:
#   node /tmp/codex-autoresearch-src/plugins/codex-autoresearch/scripts/autoresearch.mjs setup --cwd <project> --name <s> --metric-name seconds --direction lower --benchmark-command "<cmd printing METRIC name=value>" --checks-command "<cmd>"
# Intended install path is Codex's /plugins picker (TheGreenCedar -> codex-autoresearch), which bundles the CLI.

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

Commands — how to trigger Codex Autoresearch

/codex-autoresearch Measured, resumable optimization loop with baseline, correctness gate, and evidence ledger

It also activates on plain-language prompts like these:

Start a new Codex Autoresearch loop
Resume and inspect the current autoresearch run
Show the autoresearch dashboard and quality gaps

Frequently asked questions

Is the Codex Autoresearch skill free?: Yes. The skill itself is free from TheGreenCedar/codex-autoresearch. SkillProof publishes the install command and an independent test verdict at no cost.
Does Codex Autoresearch work with Claude Code?: We tested it with Claude Code 2.x (agent harness) on Jul 21, 2026. Verdict: Tested · Works. I built a toy Node project (sumTo via array+reduce, a bench.mjs printing METRIC seconds= and a checks.mjs), then ran the loop two ways under a temp HOME. Baseline (no skill): I just rewrote the function to an accumulator loop and re-ran bench (0.7297s -> 0.3652s) — a faster file, no record. Skill: setup wrote a session (autoresearch.md/.jsonl/.sh), doctor --check-benchmark confirmed the METRIC contract, next+log captured baseline 0.7234, then the keep packet logged 0.3542 with checks passing (exit 0, "checks OK"), created scoped git commit 8d4e192 touching only lib/sum.mjs, attached an ASI note (hypothesis/evidence/next-action), and finalize-preview produced a review-branch plan flagging maturity "experimental". Decisive difference: when I fed a faster-but-wrong change (closed-form +1), the packet's checks exited 1 and the CLI forced status checks_failed as the only allowed status, refusing the keep — the baseline path has no such regression gate.
What is the Codex Autoresearch SkillProof Score?: 8.8/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 8/10, docs & honesty 4/5.
How do I install Codex Autoresearch?: 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 Codex Autoresearch 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.