Skip to main content
skill CX Codex

autoresearch · factory-ai — Claude Code skill: what it does, how to install

The autoresearch skill runs repeated code experiments against a measurable target. Use it for tuning tests, builds, bundle size, or ML training results.

no license 137 installs ★ 129 updated 2026-10-25

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add Factory-AI/factory-plugins --skill autoresearch -a claude-code
Open on GitHub

What autoresearch does

autoresearch sets up a repeatable optimization session around a command and a named metric. The agent changes files in scope, runs the measurement, compares the result with the current best, and keeps or reverts the change. It also supports a termination rule such as an experiment count, time budget, target value, or interruption.

The setup uses a Git branch plus autoresearch.md, autoresearch.sh, and autoresearch.jsonl. An optional autoresearch.checks.sh handles required tests or type checks. The log records experiment outcomes, while MAD-based confidence scoring helps assess noisy measurements. The loop cannot start without a measurable metric.

What it helps with

  • define a measurable optimization target
  • record a baseline with autoresearch.sh
  • compare trials against the current best result
  • revert changes that worsen the metric
  • keep hypotheses and outcomes in JSONL

Example prompts for the agent

  • Set up autoresearch to reduce the runtime reported by pnpm test and limit the editable files.
  • Create autoresearch.sh, measure the baseline, and log it in autoresearch.jsonl.
  • Run another experiment, compare its metric with the best result, and revert the change if it regresses.

How to install autoresearch

1Skills CLI (Vercel)

npx skills add Factory-AI/factory-plugins --skill autoresearch -a claude-code

2Manually: put the files in place

~/.claude/skills/autoresearch/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/autoresearch/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about autoresearch

How do I install autoresearch in Claude Code?

Run npx skills add Factory-AI/factory-plugins --skill autoresearch -a claude-code, then check ~/.claude/skills/autoresearch/SKILL.md.

What does autoresearch do?

It structures an optimization loop around code edits and measurements. Successful changes stay in the branch, while regressions are reverted and recorded.

Which files does the experiment setup create?

The core files are autoresearch.md, autoresearch.sh, and autoresearch.jsonl. Add autoresearch.checks.sh when tests or type checks are part of the constraints.

Similar skills

All skills for Agent setup: memory, context, orchestration →

autoresearch by other authors comes first.

autoresearch official luiscantero autoresearch — a GitHub Copilot skill that runs an autonomous experiment loop: it edits code, tests, measures a metric, and keeps only what improves it. The project must be a git repository with terminal access. CX Codex +1 3,400installs autoresearch uditgoenka The autoresearch skill changes code, checks it against a metric, then keeps or discards the result. Separate commands cover debugging, security audits, and other workflows. CX Codex +1 2,700installs autoresearch Yeachan-Heo autoresearch runs one mission through repeated change-and-evaluate cycles, writing each result to JSON and a markdown decision log. Useful when the work needs a long recorded search rather than one attempt. CX Codex +1 654installs autoresearch Orchestra Research autoresearch manages autonomous research projects using a two-loop cycle for rapid experiment execution and periodic synthesis. CX Codex +1 1,200installs autoresearch AgriciDaniel autoresearch runs a bounded research loop with web access, drafts a cited report, and lets you review before filing results. Use when you need deep investigation with fact-checking built in. CX Codex +1 572installs autoresearch-agent Alireza Rezvani autoresearch-agent runs repeated file experiments against a measurable metric, keeping gains and reverting failures. Use it for code, tests, prompts, or content. CX Codex +1 589installs dueling-autoresearch gaasher dueling-autoresearch pits two distinct technical approaches against each other on a shared metric. It runs autonomous research loops until interrupted. CX Codex +1 ★ 185repo stars skill-autoresearch akillness skill-autoresearch runs edits to skills, prompts and SOPs through a frozen local benchmark: one change per run, then keep or revert. CX Codex +1 323installs ar-run Alireza Rezvani ar-run runs exactly one autoresearch iteration: reviews results history, modifies a target file, and evaluates metric improvements. CX Codex +1 ★ 30,251repo stars tournament-autoresearch gaasher The tournament-autoresearch skill pits architecture proposals against each other before an ML experiment runs. It suits projects where you want an agent to choose one idea and learn from its measured result. CX Codex +1 ★ 185repo stars arbor K-Dense-AI arbor — a skill for Claude Code, Codex, Cursor, GitHub Copilot and Gemini CLI that runs an autonomous optimization loop over a persistent hypothesis tree. CX Codex +1 1,500installs

From the same set17

The whole set →
threat-model-generation threat-model-generation — a skill for Claude Code, Codex, Cursor, and other agents that builds a STRIDE threat model from a repository's codebase. CX Codex +1 151installs ban-type-assertions ban-type-assertions is a Claude Code skill that blocks TypeScript as assertions and guides their replacement with compiler-checked code. It runs when a package adopts the ban or needs existing lint violations fixed. CX Codex +1 151installs fix-knip-unused-exports fix-knip-unused-exports resolves knip “Unused exports” violations by removing or relocating exports that have no production consumers. Use when npm run knip reports unused code. CX Codex +1 147installs skill-creation skill-creation converts past session breakthroughs into reusable agent skills inside SKILL.md files. Helps extract repeatable patterns and refine existing instructions. CX Codex +1 142installs vulnerability-validation vulnerability-validation is a Claude Code skill that tests whether security scan findings can be exploited and filters false alarms. It runs after commit-security-scan, before findings become issues or block a PR. CX Codex +1 135installs commit-security-scan The commit-security-scan skill reviews Git changes for security flaws using threat-model context. Use it for pull requests, staged files, or branch diffs. CX Codex +1 126installs http-toolkit-intercept http-toolkit-intercept captures HTTP requests and responses from any script or service — including LLM API calls, backend requests, and auth flows. Use it when you need to see what's actually sent over the wire and what comes back. CX Codex +1 124installs browser-navigation browser-navigation automates browser tasks: page navigation, form filling, screenshots, data extraction, and session recording. CX Codex +1 120installs

Project sponsors

Companies that keep this analytics open