research-refine-pipeline

Chains method refinement and experiment planning into one end-to-end research proposal workflow.

Updated Jul 9, 2026
One-click install
npx skills add https://github.com/Lingjie-wang/autoRL --skill research-refine-pipeline-lingjie-wang
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: research-refine-pipeline
Source: https://github.com/Lingjie-wang/autoRL/tree/main/Auto-claude-code-research-in-sleep/skills/research-refine-pipeline
Command: npx skills add https://github.com/Lingjie-wang/autoRL --skill research-refine-pipeline-lingjie-wang

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning a vague research direction into both a focused method proposal and a detailed experiment roadmap normally requires running separate workflows and manually keeping them consistent. This Skill chains the research-refine and experiment-plan stages so the final proposal, review history, and validation plan stay aligned in one pass. ## Core Features & Use Cases - End-to-End Pipeline: Runs method refinement first, then generates a claim-driven experiment plan grounded in the refined proposal. - Planning Gate: Enforces a checkpoint that verifies the method thesis, dominant contribution, and rejected complexity before any experiment planning begins. - Structured Outputs: Produces FINAL_PROPOSAL.md, REVIEW_SUMMARY.md, EXPERIMENT_PLAN.md, EXPERIMENT_TRACKER.md, and a PIPELINE_SUMMARY.md under refine-logs/. - Use Case: A researcher with a rough idea for a paper asks to "串起来" the whole flow and receives a stabilized method thesis plus a prioritized experiment roadmap with run order, budget, and decision gates. ## Quick Start Ask the assistant to run the research-refine-pipeline on your rough research idea to produce a refined proposal and experiment plan together.

Frequently Asked Questions about research-refine-pipeline

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

FAQPage Schema
How do I go from a vague research idea to an experiment plan?

Run the research-refine-pipeline with your rough direction as the argument. It first stabilizes the method thesis through the research-refine stage, then generates a claim-driven experiment plan with run order, budget, and decision gates.

What is the difference between research-refine and research-refine-pipeline?

research-refine only handles method refinement and stops at the final proposal. research-refine-pipeline chains that stage with experiment-plan, so you also get an experiment roadmap and a pipeline summary in one run.

What output files does the research pipeline generate?

It writes FINAL_PROPOSAL.md, REVIEW_SUMMARY.md, REFINEMENT_REPORT.md, EXPERIMENT_PLAN.md, EXPERIMENT_TRACKER.md, and PIPELINE_SUMMARY.md under the refine-logs directory, with timestamped versions logged to a manifest.

Can I reuse an existing proposal instead of re-running refinement?

Yes. The triage phase checks whether refine-logs/FINAL_PROPOSAL.md already exists and still matches the current request. If it is strong and aligned, the pipeline skips refinement and jumps directly to experiment planning.

When should I not use the full research pipeline?

Use the staged skills instead when you only need one stage: research-refine for method refinement only, or experiment-plan for validation planning on an already stable proposal. The pipeline is meant for the integrated one-shot flow.