research-engineer

Runs the research loop from paper understanding through implementation, experiments, and analysis.

2|1|Updated Mar 7, 2026
One-click install
npx skills add https://github.com/AutoByteus/autobyteus-agents --skill research-engineer-autobyteus
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: research-engineer
Source: https://github.com/AutoByteus/autobyteus-agents/tree/main/agents/research-engineer/skills/research-engineer
Command: npx skills add https://github.com/AutoByteus/autobyteus-agents --skill research-engineer-autobyteus

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Research engineering work—reproducing a paper, implementing a method, running experiments, and interpreting results—often loses track of assumptions, sources, and failed runs. This Skill structures that entire loop so every decision, experiment, and conclusion stays traceable and evidence-grounded. ## Core Features & Use Cases - Full Research Loop: Covers source discovery, literature ranking, paper analysis, research planning, minimal implementation, experiment design, execution, debugging, and self-review. - Structured Artifact Templates: Produces research briefs, notes, source indexes, literature reviews, paper analyses, experiment plans and reports, loop logs, and decision logs from ready-made templates. - Honest Evidence Handling: Preserves negative results, separates paper-faithful reproduction from exploratory variants, and matches conclusion strength to evidence strength. - Use Case: Ask it to reproduce an attention mechanism from a paper: it will analyze the paper, plan a minimal implementation, run smoke tests and training experiments, log each loop, and deliver an experiment report with a clear next-step decision. ## Quick Start Use the research-engineer skill to reproduce the method from this paper, run a small validation experiment, and report whether the results support the paper's claims.

Frequently Asked Questions about research-engineer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reproduce a machine learning paper with an AI agent?▼

Provide the paper or link and state your constraints. The skill analyzes the method, writes a paper analysis and research plan, builds a minimal faithful implementation, runs smoke tests and experiments, and delivers an experiment report with reproducibility details.

What artifacts does a research engineering workflow produce?▼

It produces a research brief, continuous research notes, a source index, literature review, paper analysis, research and experiment plans, an implementation handoff, a loop log, an experiment report, and a decision log, plus an optional HTML explainer.

Can it run training or fine-tuning experiments locally?▼

Yes. When the plan requires empirical validation, it handles environment setup, data preparation, training or fine-tuning runs, checkpoint tracking, metric evaluation, and debugging, recording commands, seeds, and raw output paths.

Does it work without internet access or only with supplied sources?▼

Yes. It respects user constraints on source gathering and can operate in supplied-source-only mode, recording access limits in the source index and justifying any gaps in source discovery.

How does it handle failed experiments or negative results?▼

Failed runs, unstable training, and mismatched reproduction results are preserved as research evidence in the loop log and experiment report, with interpretation, confounders, and a concrete next-loop decision rather than being discarded.

When should I use this instead of a paper-reading assistant?▼

Use it when the task goes beyond reading and Q&A into implementation, reproduction, training, benchmarking, or empirical validation. For pure paper discovery and evidence-aware question answering, a narrower paper research assistant fits better.