autoresearch-karpathyAutonomous AI research skill for running automated neural network experiments. This skill should be used when the user wants to set up autonomous AI research...
Install via ClawdBot CLI:
clawdbot install baiyunrei2025/autoresearch-karpathyGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Potentially destructive shell commands in tool definitions
eval(Calls external URL not in known-safe list
https://github.com/karpathy/autoresearchAI Analysis
The skill's primary risk is executing arbitrary code modifications and shell commands in an autonomous research loop, which could lead to unintended system changes or resource exhaustion. It clones and runs an external repository, but this is consistent with its stated purpose of automating AI experiments. No evidence of data exfiltration, credential harvesting, or hidden malicious instructions was found in the provided definition.
Audited Apr 16, 2026 · audit v1.0
Generated May 9, 2026
Run an autonomous loop to test thousands of hyperparameter configurations overnight. Each experiment runs for 5 minutes, logs results, and keeps only improvements. Wake up to a log of experiments and a better model.
Let the agent modify the model architecture, such as depth, width, and activation functions, to find optimal configurations for language modeling. The agent autonomously explores and validates architectures.
Experiment with different optimizers (Muon, AdamW) and their parameters (learning rate, weight decay) to minimize validation bits per byte. The agent autonomously iterates and discards bad runs.
Deploy the skill for overnight or multi-day runs where the agent continuously modifies training code, runs experiments, evaluates results, and improves the model without human intervention.
Offer a cloud-based platform where users can submit their model training code and let an AI agent autonomously optimize it. Revenue from subscription tiers based on GPU hours and experiment volume.
Provide a service to clients who want to optimize their neural network models. Use the skill to run autonomous experiments and deliver improved model configurations. Charge per project or per model improvement.
Bundle GPU compute with the autonomous research skill, allowing customers to run experiments without managing infrastructure. Revenue from compute margins and a markup for the AI optimization service.
💬 Integration Tip
Ensure you have a single NVIDIA GPU (e.g., H100) and Python 3.10+ with the uv package manager. The skill runs on a git branch and requires careful initial setup of the autoresearch repository.
Scored May 9, 2026
Multi-source deep research agent. Searches the web, synthesizes findings, and delivers cited reports. No API keys required.
Transparent, rigorous research with full methodology — not a black-box API wrapper. Conducts exhaustive investigation through mandated 2-cycle research per t...
Seven-stage deep research pipeline for substantive research questions. Activates when user asks for deep research, 深度研究, research report, evidence-backed inv...
Perform deep research using Claude by autonomously gathering, analyzing, and synthesizing detailed information across diverse topics on request.
全生命周期质量保障专家，覆盖需求、设计、编码、测试、上线、运维全阶段的质量保障活动。 融合缺陷预防（逆向操作、依赖踏空、并发冲突、新旧兼容、状态迁移、因果判定）与质量度量、持续改进三大支柱， 帮助团队建立「预防-评审-度量-改进」的质量闭环，实现软件质量的持续提升。 Use when: - 需求/设计/编码/测试...
专家团自动组建技能（反面教材）。核心观点：AI只认事不认人，没必要模拟人类多角色协作方式。本技能演示为什么不需要这样做——AI应该直接面对任务，而不是模拟人类团队。触发词：组建专家团、专家协作、团队完成任务、自动组建团队、expert team、专家团、团队协作完成任务。