plantclef-2026

PlantCLEF 2026

Research project exploring fine-grained visual categorization in plant species identification using foundation models.

Quick Start

Repository Structure

plantclef-2026/
├── skills/                     # Agent Skills (auto-discovered by compatible agents)
│   ├── pace-cluster/           # PACE Phoenix cluster operations
│   ├── experiment/             # Experiment lifecycle management
│   └── research-docs/          # Research documentation workflow
├── .claude/                    # Claude Code config & slash commands
│   ├── settings.json
│   └── commands/
│       ├── docs/               # /docs:concept, /docs:reference, /docs:vendor
│       ├── experiment/         # /experiment:proposal, /experiment:result, /experiment:validate
│       └── openspec/           # OPSX commands (empty until you run `openspec init`)
├── docs/
│   ├── README.md               # Documentation index
│   ├── guides/                 # How-to documentation
│   ├── _templates/             # Document templates (don't modify)
│   ├── concepts/               # Research ideas
│   ├── references/             # Papers and AI research outputs
│   └── vendor/                 # External tool documentation
├── openspec/
│   ├── config.yaml             # Project context and schema
│   ├── specs/                  # Built specifications
│   └── changes/                # Change proposals
├── user/
│   └── murilo/                 # User workspace
│       └── experiments/
│           └── 000-template/   # Template experiment
│               └── pyproject.toml  # Experiment-level dependencies
├── vendor/                     # Vendored submodules / external repos
├── pyproject.toml              # Workspace root (discovers experiments)
├── AGENTS.md                   # AI assistant guidelines
├── CLAUDE.md                   # Symlink → AGENTS.md
├── ruff.toml                   # Ruff linter/formatter config
├── .pre-commit-config.yaml     # Pre-commit hooks
├── .gitignore
└── LICENSE

Setup

1. Install uv

curl -LsSf https://astral.sh/uv/install.sh | sh

2. Create Your User Workspace

Each team member has their own workspace under user/<name>/:

cp -r user/example user/<your-name>

3. Set Up Your Environment

Edit the pyproject.toml inside your first experiment (e.g., user/<your-name>/experiments/001-my-experiment/pyproject.toml) with your name and dependencies, then:

# Create virtual environment
uv venv .venv

# Activate it
source .venv/bin/activate

# Install experiment dependencies
uv sync --package <experiment-name>

The root pyproject.toml defines a uv workspace that auto-discovers all experiments under user/*/experiments/*. Each experiment has its own pyproject.toml for dependency isolation — different experiments can use different library versions.

4. Install Pre-commit Hooks

pre-commit install

Agent Tooling

This repository uses three systems for AI-assisted workflows. See Agent Tooling Guide for full details.

Skills (Auto-Activated)

Agent Skills are auto-discovered by compatible agents (Claude Code, Cursor, Gemini CLI, etc.). Just describe what you need and the agent loads the right skill.

Slash Commands (User-Invoked)

Claude Code-specific shortcuts for explicit workflow triggers:

OpenSpec (Change Management)

Spec-driven development using the OPSX workflow. Run openspec init after cloning to generate skills and commands. See OpenSpec Guide.

Documentation

See docs/README.md for the full documentation index.

Contributing

1. Create your user workspace under user/<your-name>/
2. Use slash commands to create structured documents
3. Follow the research workflow: Reference → Concept → Experiment → Results
4. See AGENTS.md for AI assistant conventions