Autoresearch Agent
Autonomous experiment loop that optimizes any file by a measurable metric.
Before & after
Same prompt, same agent — the only difference is the skill.
I can help you optimize your homepage copy to increase email signups. To set this up, I'd need to understand your current performance baseline. How many visitors do you get daily, and what's your current signup conversion rate? Also, what aspect of the copy are you most interested in optimizing—the headline, the value prop, the CTA? I could manually rewrite variations and test them, but it might be more efficient to set up an automated optimization loop. Do you have analytics set up to measure conversions per variant? Once you clarify, I could start iterating on the copy.
AUTORESEARCH AGENT: HOMEPAGE COPY OPTIMIZATION SETUP Target file: /app/copy/homepage.md Baseline: • Current headline: Ship marketing campaigns faster • Current CTA: Start free trial • Current conversion: 3.2% (signups / visitors) Measurement: Google Analytics event signup_click / page_view Optimization metric: Conversion rate (higher = better) Iteration limit: 20 attempts, stop if no improvement > 0.1% in last 5 attempts ===== EXPERIMENT LOOP ===== ROUND 1 Baseline metric: 3.2% Iteration: Edit headline • Previous: Ship marketing campaigns faster • New: Get campaign ideas that actually convert • Reasoning: Benefit-focused, specificity Commit: git add -A && git commit -m "Experiment: benefit-focused headline" Evaluate: Run script → A/B test deployed → 2,000 visitors served → Signups: 68 (3.4% conversion) → Delta: +0.2pp Result: PASS (+0.2pp improvement) → Keep change (don't reset) ROUND 2 Baseline metric: 3.4% (new) Iteration: Edit subheading • Previous: Less time on setup, more on strategy • New: 100+ marketers generate campaigns daily. Yours could be next. • Reasoning: Social proof + urgency Commit: git add -A && git commit -m "Experiment: social proof subheading" Evaluate: → Signups: 71 (3.55% conversion) → Delta: +0.15pp Result: PASS (+0.15pp) → Keep change ROUND 3 Baseline metric: 3.55% Iteration: Edit CTA copy • Previous: Start free trial • New: See it work (free, 5 minutes) • Reasoning: Reduce friction (show time commitment) Evaluate: → Signups: 69 (3.45% conversion) → Delta: -0.1pp Result: FAIL (worse) → Reset via git reset --hard HEAD~1 → Discard change → Back to 3.55% baseline ROUND 4 (Agent learns from failure) Iteration: Edit CTA + subheading together • CTA: Try free (shorter, past failure informed this) • Subheading: Join teams cutting campaign setup time by 70% Evaluate: → Signups: 76 (3.8% conversion) → Delta: +0.25pp (bigger win!) Result: PASS (+0.25pp) → Keep ===== METRICS AFTER 4 ROUNDS ===== Baseline: 3.2% Final: 3.8% Total improvement: +0.6pp (+18.75% relative lift) Cumulative conversions gained (extrapolated): 200 extra signups/month at 10K monthly visitors ===== AUTOMATION ===== Git-tracked experiments: • Each round = 1 commit • Commit message = hypothesis • Metric = part of commit message (for tracking) git log --oneline: • Experiment: social proof subheading (+0.15pp) • Experiment: benefit-focused headline (+0.2pp) • Baseline: homepage copy 3.2% baseline Failure handling: • If eval script errors → abort round, log error, retry next round • If metric goes negative (error rate) → git reset --hard, alert ===== WHAT'S BEING OPTIMIZED ===== Each round edits ONE section (or 1–2 related elements): 1. Headline 2. Subheading 3. CTA (copy) 4. Value proposition 5. Social proof 6. FAQ Agent reasons about each: Social proof helped before, try it on CTA? ===== DEPLOYMENT ===== • Experimental changes deployed to production immediately • Each variant shown to 5% traffic (95% gets baseline) • After 2,000 visitors (statistical significance), promote winner or revert • All wins persist (don't reset to baseline each round)
About this skill
name: autoresearch-agent description: Use when Autonomous experiment loop that optimizes any file by a measurable metric.
Autoresearch Agent
Autonomous experiment loop that optimizes any file by a measurable metric. Inspired by the original author's autoresearch. The agent edits a target file, runs a fixed evaluation, keeps improvements (git commit), discards failures (git reset), and loops indefinitely. Use when: user wants to optimize code speed, reduce bundle/image size, improve test pass rate, optimize prompts, improve content quality (headlines, copy, CTR), or run any measurable improvement loop. Requires: a target file, an evaluation command that outputs a metric, and a git repo.
What you get
- Public GitHub repo (alirezarezvani/claude-skills)
- the autoresearch-agent skill folder with SKILL.md. Part of a 337-skill / 30-agent / 70-command install.
Customize your output
- Fork the repo and adapt the skill's instructions and references to your workflow.
Example output
Activates automatically when your request matches Autoresearch Agent; chains with the other skills, agents, and commands in the collection.
Best for
Creators, builders, and teams using Claude Code.
SKILL.md preview
---
name: autoresearch-agent
description: Use when you need an autonomous loop that repeatedly edits a target file, evaluates it against a measurable metric, and keeps only improvements.
version: 1.0.0
category: Development / Engineering
author: AgentVolt
license: proprietary
tags:
- development
- engineering
---
# Autoresearch Agent
Runs an autonomous loop against a single target file: edit, evaluate against a fixed metric, keep improvements via git commit, discard failures via git reset, repeat.
## When to use
… (sign up to view the full skill)More development skills
View all Development skills →Generate
Scaffold code, tests, and boilerplate from a short spec.
Performance Profiler
Systematic performance profiling for Node.js, Python, and Go applications.
Threat Detection
Use when hunting for threats in an environment, analyzing IOCs, or detecting behavioral anomalies in telemetry.
Agent Harness
Turns a domain folder of skills into a bounded agentic loop: compile a goal into a task plan, execute with the domain's tools, verify every task, retry within a budget.