jbenhart44

Research Toolkit — AI skill for Claude Code

AI community

Reusable commands for structured AI-assisted research workflows.

How to install Research Toolkit

This entry records only its repository, not the path inside it, so there is no exact command to give. Open jbenhart44/Research-Toolkit and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Research Toolkit does

Reusable commands for structured AI-assisted research workflows. Plan-Construct-Verify (PCV), Council of Agents (CoA), Parallel Agent Consensus Engine (PACE), and workflow automation. Built and validated with real research sessions using Claude Code during research pursuing a PhD in Operations Research.

Alternatives in AI

Open Multi Agent — TypeScript multi-agent orchestration engine — one runTeam() call from goal to result 5.8k ★
Golutra — Multi-agent AI orchestration platform for automation, workflows, and developer tools 3.8k ★
Define — Definition phase - Clarify and scope problems with multi-AI consensus 2.8k ★

README

Research Amp Toolkit

*Every empirical value cites a file on disk. Every claim survives `grep`. Dissertation-grade, not vibe-grade.*

**⚠️ v0.1 — experimental, no support commitment.** This is an early research release under active development. Breaking changes expected between versions. Issues and pull requests are welcome; responses are best-effort, not guaranteed. If you're looking for a stable production tool, this isn't it yet.

15 Claude Code commands that amplify your research. The tools don't think — they verify, structure, and document. You bring the signal. Validated across hundreds of production sessions.

**Authors:** Jake Benhart & Dr. Michael G. Kay (NC State University — Operations Research)

**Links:** [Landing page](https://jbenhart44.github.io/Research-Toolkit/) · [Download v0.1](https://github.com/jbenhart44/Research-Toolkit/releases) · [Report an issue](https://github.com/jbenhart44/Research-Toolkit/issues) · [Discussions](https://github.com/jbenhart44/Research-Toolkit/discussions) · [Contribute](CONTRIBUTING.md)

Design rationale: [DESIGN.md](DESIGN.md). Who this is for and not for: [POSITIONING.md](POSITIONING.md). Error classes the toolkit is built to prevent: [references/preventable_errors.md](references/preventable_errors.md). Consolidated rule index: [references/iron_rules.md](references/iron_rules.md).

Spec-Driven Development

This toolkit is a working implementation of **Spec-Driven Development (SDD)** — the practice of writing the plan, constructing against it, and verifying the result as three separate, auditable steps rather than one conflated "just ask the model" loop. The commands here (`/pcv`, `/pace`, `/audit`, `/readable`, `/coa`) predate the SDD label; they grew out of PhD-research needs where a wrong citation or a hallucinated number is a career cost, not a minor bug. The SDD label is retroactive — the pattern became visible when the broader AI-engineering conversation converged on the same decomposition (peer implemen