Run the full AI research pipeline from your terminal — no browser needed.
# 1. Add the marketplace source
/plugin marketplace add OpenLAIR/dr-claw-plugin-cc
# 2. Install the plugin
/plugin install dr-claw@dr-claw
# 3. Reload
/reload-plugins# Clone the repo
git clone https://github.com/OpenLAIR/dr-claw-plugin-cc.git ~/dr-claw-plugin-cc
# Launch Claude Code with the plugin
cc --plugin-dir ~/dr-claw-plugin-cc/plugins/dr-clawOnce accepted into claude-plugins-official:
/plugin install dr-claw@dr-claw
cd ~/my-research-project
/drclaw:setup # Interactive wizard — asks topic, venue, hypothesis
/drclaw:status # View pipeline progress
/drclaw:run # Execute the next task
/drclaw:run all # Execute all remaining tasks
| Command | Description |
|---|---|
/drclaw:setup |
Initialize a research pipeline in the current directory |
/drclaw:status |
Show pipeline progress across all 5 stages |
/drclaw:run [next|all|task-id] |
Execute next task, all tasks, or a specific task |
/drclaw:reset [all|stage|task-id] |
Reset task statuses back to pending |
Survey → Ideation → Experiment → Publication → Promotion
Each stage has tasks with skill references and dependency chains. Tasks execute sequentially, respecting dependencies.
/drclaw:setup supports three research templates:
- Method/Model — 13 tasks for algorithmic or architectural innovation
- Dataset — 14 tasks for benchmark creation or data curation
- Position Paper — 13 tasks for surveys or perspective papers
.pipeline/
├── docs/research_brief.json # Your project definition
├── tasks/tasks.json # Task list with statuses
├── Survey/reports/ # Literature reviews, gap analysis
├── Ideation/ideas/ # Generated and evaluated ideas
├── Experiment/core_code/ # Implementation code
├── Experiment/analysis/ # Results, figures, tables
├── Publication/paper/ # Paper draft, review reports
└── Promotion/slides/ # Presentation materials
60+ research skills covering the full pipeline:
- Survey: deep-research, academic-researcher, dataset-discovery
- Ideation: idea-generation, idea-evaluation, resource-preparation
- Experiment: code-survey, experiment-dev, experiment-analysis
- Publication: paper-writing, figure-gen, paper-reviewer, reference-audit
- Promotion: academic-presentations
- Infrastructure: distributed-training, inference-serving, RAG, fine-tuning, and more
When you open Claude Code in a directory with .pipeline/, the plugin automatically detects it and shows your current progress — no need to run /drclaw:status manually.