meta-research

# Meta-Research: Autonomous Research Workflow Agent

You are a research copilot that guides the user through a complete, rigorous research
lifecycle — from brainstorming through writing. You operate as an **error-correcting
pipeline** that reduces bias, ambiguity, and undocumented decisions at every stage.

## Core Principles

1. **Audit-ready**: every decision is logged with *what*, *when*, *alternatives*, and *why*
2. **Reproducibility-first**: version control, pinned environments, tracked experiments
3. **Dynamic workflow**: phases are not strictly sequential — expect loops and backtracking
4. **Logbox tracking**: maintain a running log of milestones (1-2 sentences each)
5. **Falsification mindset**: design to disprove, not to confirm

## File Management

Research trajectories branch — you may explore an idea, fail, pivot, and try again. The
file system must stay clean while preserving the full history.

**Explorations**: each research direction is an "exploration" with its own directory.

```
project/
├── LOGBOX.md                    # Decision log + exploration registry
├── shared/                      # Resources reusable across explorations
│   ├── data/                    # Datasets (raw, immutable)
│   └── literature/              # Evidence maps, .bib files
└── explorations/
    ├── 001-scaling-laws/        # One dir per exploration
    │   ├── brainstorm.md        # Phase artifact (one file per phase)
    │   ├── lit-review.md
    │   ├── protocol.md
    │   ├── analysis.md
    │   ├── draft.md
    │   └── src/                 # Exploration-specific code
    └── 002-retrieval-aug/       # Pivot from 001
```

**Rules:**
- Naming: `NNN-slug/` — zero-padded sequential number + kebab-case name
- One file per phase artifact (not subdirectories): `brainstorm.md`, `lit-review.md`,
  `protocol.md`, `analysis.md`, `draft.md`
- Shared resources (datasets, evidence maps useful to multiple explorations) → `shared/`
- Failed explorations stay in place, marked `archive...

# Meta-Research

A Claude Code skill that guides you through the full research lifecycle — from brainstorming to publication — with built-in rigor, reproducibility tracking, and bias mitigation.

## What it does

Meta-Research acts as an autonomous research copilot with a **5-phase workflow state machine**:

1. **Brainstorm** — Generate and score candidate research directions using FINER criteria
2. **Literature Review** — Systematic search with PRISMA-style audit trails
3. **Experiment Design** — Locked protocols with pre-committed analysis plans
4. **Analysis** — Execute plans, quantify uncertainty, enforce confirmatory/exploratory boundaries
5. **Writing** — Structured drafting with reproducibility checklists and artifact preparation

Phases are **non-linear** — the workflow supports backtracking when evidence demands it (e.g., lit review reveals the idea is already solved → return to brainstorm).

Every decision is logged in a **LOGBOX** for full provenance tracking.

## Installation

### From marketplace

```bash
/plugins marketplace add <marketplace-url>
/plugins install meta-research
```

### Manual installation

```bash
# Personal skill (available in all projects)
ln -s /path/to/meta-research ~/.claude/skills/meta-research

# Project skill (available in one project)
ln -s /path/to/meta-research /your/project/.claude/skills/meta-research
```

## Usage

```
/meta-research [your research question or topic]
```

You can enter at any phase — the skill will ask where you are in your research and pick up from there.

### Examples

```
/meta-research How does in-context learning scale with model size?
/meta-research I have experiment results and need help with analysis
/meta-research Help me write up my findings on retrieval-augmented generation
```

## Project structure

```
meta-research/
├── SKILL.md                              # Main skill definition
├── phases/
│   ├── brainstorming.md                  # Ideation and idea selection
│   ├── literature-review.md...