agentpluginsdirectory
RDCrystalLab avatar

autoresearch

v1.3.0

by RDCrystalLab · author: Udit Goenka

autoresearch turns a coding agent into a loop: change something, run the check, keep the change if the metric improved, discard it if not, repeat. It follows Karpathy's autoresearch idea of one constraint plus one mechanical metric. Udit Goenka wrote it; RDCrystalLab packaged it as an Agent Plugin. Before the loop starts it reads the files in scope, asks for the metric, and records a baseline as iteration zero.

Autonomous improvement engine for Claude Code. Runs an unbounded modify-verify-keep/discard loop against any mechanical metric. Includes planning wizard, security audit, shipping workflow, autonomous debugger, and autonomous fixer.

What autoresearch does, in the publisher's words

Turn Claude Code into a relentless improvement engine.

Based on Karpathy's autoresearch: constraint + mechanical metric + autonomous iteration = compounding gains.

"Set the GOAL → Claude runs the LOOP → You wake up to results"

  • Read context: reads all in-scope files
  • Define goal: extracts or asks for a mechanical metric
  • Define scope: which files can be modified vs read-only
  • Establish baseline: runs verification on current state (iteration #0)
  • Confirm and go: shows setup, then begins the loop

From the project README, punctuation lightly normalized · Open the README on GitHub

Skills
1
MCP servers
0
Stars
0
License
n/a
Repo created
2026-01-10
Last pushed
2026-08-15
Publisher type
Organization
Version
1.3.0

Links

Skills · 1

autoresearch
自主目標導向迭代（Autonomous Goal-directed Iteration）。將 Karpathy 的 autoresearch 原則應用於任何任務。自主循環執行, 修改、驗證、保留/捨棄、重複。支援透過 Claude Code 的 /loop 指令指定可選的迴圈次數。

Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.

Category

Agent Tooling. Assigned by this directory. Agent Plugins 1.0.0 has no category field, so no manifest declares one.

Keywords

autonomous · research · improvement · debug · workflow

Related by publisher and keywords

  • build-standalone-html-reportRDCrystalLab

    建立與修訂自包含的單檔 HTML 研究、事故、架構比較與決策報告，內建編輯方法、視覺規範、範本與驗證工具。

  • gh-toolsRDCrystalLab

    GitHub workflow commands for issues, PRs, and project management using the gh CLI

  • agency-agentsRDCrystalLab

    數位行銷代理商專用 Agent 集合，涵蓋工程、設計、行銷、銷售、測試等 14 個專業類別，共超過 140 個專業 Agent。

  • anvilRDCrystalLab

    Evidence-first coding agent. Verifies before presenting. Attacks its own output with adversarial multi-model review, IDE diagnostics, and SQL-tracked verification to ensure code quality.

  • cavemanRDCrystalLab

    Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.

  • cli-creatorRDCrystalLab

    從 API 文件、OpenAPI spec、curl 範例或現有腳本建立可組合的 CLI 工具。支援 Python (uv single-file)、Rust、TypeScript，含 companion skill 自動生成。

Add the verified badge to your README[![Verified by AgentPluginsDirectory.com against Schema 1.0.0](https://agentpluginsdirectory.com/badge.svg)](https://agentpluginsdirectory.com/plugins/autoresearch)

✓ Verified . We fetched the manifest from GitHub and checked it against the official Agent Plugins 1.0.0 schema at agent-plugins.org.