# uditgoenka/autoresearch

Claude Autoresearch Skill — Autonomous goal-directed iteration for Claude Code. Inspired by Karpathy's autoresearch. Modify → Verify → Keep/Discard → Repeat forever.

Repository: https://github.com/uditgoenka/autoresearch
Canonical: https://ross.abutalabs.com/products/uditgoenka-autoresearch
Homepage: https://udit.co/projects/autoresearch
Language: Shell
License: MIT
License Family: permissive
Topics: ai, autonomous-agent, autoresearch, claude, claude-code, iteration, karpathy, productivity, skill
Last push: 2026-08-12T22:28:18+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 97, release rhythm 97, longevity 12
- inputs: {"age_days": 173, "days_push": 21, "days_rel": 21, "gap_med": 0, "n_releases_24m": 44}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5945, forks 451 (observed 2026-08-28T04:09:32.650144+00:00)

## What it is
A Claude Code skill (also compatible with OpenCode and OpenAI Codex) that turns coding agents into autonomous goal-directed iteration loops: modify, verify a mechanical metric, keep or discard via git rollback, and repeat. Inspired by Karpathy's autoresearch, it includes an orchestrator, bounded loops, and security-audit subcommands.

## Use cases
- run autonomous improvement loops on a codebase overnight
- increase test coverage to a target percentage automatically
- automatically find and fix bugs with verification
- run STRIDE/OWASP security audits iteratively with auto-fix
- optimize a metric like build time or lighthouse score without human input
- set a goal and metric and let the agent iterate until done

## When to choose
- you use Claude Code, OpenCode, or Codex and want autonomous, metric-driven iteration
- you want atomic changes with automatic git rollback on failure
- you need a repeatable loop for coverage, performance, refactoring, or security hardening

## When to avoid
- you need human review of every change before it's committed
- your task has no mechanical, scriptable success metric
- you don't use any supported agentic coding CLI

## Facets
- artifact type: plugin
- maturity: active
- function: agent-framework, workflow-automation, developer-tools, testing, security
- domain: developer-tools
- platform: cli, cross-platform
- tags: claude-code, claude-code-skill, autonomous-iteration, opencode, openai-codex, autoresearch, git-rollback, shell, ai-agents, automation

## Member repositories
- uditgoenka/autoresearch (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:32.650144+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:50:44.861396+00:00, confidence not recorded.
  - readme: https://github.com/uditgoenka/autoresearch (fetched 2026-08-28T04:09:32.650144+00:00, sha 103d547dd08d)
  - homepage: https://udit.co/projects/autoresearch (fetched 2026-08-29T08:46:17.382132+00:00, sha 1ba6b4be419f)
  - site_page: https://udit.co/ (fetched 2026-08-29T08:46:17.390871+00:00, sha b5f1deea22df)
- Data as of 2026-08-30T08:39:29.467469+00:00.