# leo-lilinxiao/codex-autoresearch

Codex Autoresearch Skill — A self-directed iterative system for Codex that continuously cycles through: modify, verify, retain or discard, and repeat indefinitely. Inspired by Karpathy’s autoresearch concept.

Repository: https://github.com/leo-lilinxiao/codex-autoresearch
Canonical: https://ross.abutalabs.com/products/codex-autoresearch
Language: Python
License: MIT
License Family: permissive
Topics: automation, autonomous-agent, autoresearch, codex, codex-cli, codex-skill, developer-tools, iteration, openai, skill
Last push: 2026-07-13T13:53:30+00:00

## Health v2 (maintenance only)
Score: 74/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 92, release rhythm 86, longevity 12
- inputs: {"age_days": 169, "days_push": 51, "days_rel": 92, "gap_med": 19, "n_releases_24m": 6}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2255, forks 125 (observed 2026-08-28T04:06:31.287897+00:00)

## What it is
A Codex skill that runs a self-directed iterative loop: it makes one focused change, measures it against a target metric, keeps improvements, reverts failures, and repeats until the goal is reached. Inspired by Karpathy's autoresearch concept, it supports foreground and background runs with commit, verification, and rollback handled by a control script.

## Use cases
- automatically reduce test failures to zero
- iteratively improve code coverage until a target
- fix type errors and compiler warnings autonomously
- optimize latency or binary size with an agent loop
- run overnight autonomous experiments on a repository
- keep-and-revert agent-driven code changes based on measurable results

## When to choose
- you have a measurable target metric a command can evaluate
- you want autonomous, auditable iteration with automatic rollback of failed changes
- you already use OpenAI Codex CLI and want a skill-based workflow

## When to avoid
- your goal is subjective and cannot be measured by a command
- you don't use Codex or another compatible agent CLI
- you can't grant the broad permissions the loop needs to commit and revert code

## Facets
- artifact type: plugin
- maturity: active
- function: agent-framework, workflow-automation, developer-tools, testing
- domain: developer-tools, large-language-models
- platform: cli, python, cross-platform
- tags: codex-skill, autonomous-iteration, autoresearch, openai-codex, self-directed-loop, experiment-automation, git-rollback, automation, ai-agents

## Member repositories
- leo-lilinxiao/codex-autoresearch (main) score 74

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:31.287897+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 2026-08-30T02:43:45.149249+00:00, confidence not recorded.
  - readme: https://github.com/leo-lilinxiao/codex-autoresearch (fetched 2026-08-28T04:06:31.287897+00:00, sha ec8b0156b6d8)
- Data as of 2026-08-30T08:39:29.467469+00:00.