Skip to content
Get packs

Idea Discovery Pipeline

Chains literature survey, idea generation, novelty checks and review into one research-ideas workflow

Source Download skill
Repository
wanshuiyin/Auto-claude-code-research-in-sleep
Stars
18,482
Licence
MIT
Files
1 file

What it does

This skill tells an AI agent to run a multi-step research workflow: survey recent papers on a topic, brainstorm candidate ideas, check whether each idea is novel, and get critical feedback from a reviewer model (Gemini, via an MCP bridge, in this overlay version). It is written for Codex users who want Gemini as the reviewer, and it depends on several other sub-skills being installed first. The output is a ranked report in idea-stage/IDEA_REPORT.md plus a refined proposal and experiment plan for the top idea.

Great for

  • Researchers or graduate students starting from a broad topic who want a structured shortlist of testable ideas
  • ML engineers who want pilot-experiment results to guide which ideas to pursue
  • Teams already using the Auto-claude-code-research-in-sleep skill set who want a Gemini-reviewed variant

Skip it if

  • Quick one-off brainstorming: the pipeline is long, chains many sub-skills and can run pilot experiments, so it is heavy for simple questions
  • Anyone expecting verified novelty: the novelty check relies on literature search results and can miss prior work
  • Unattended runs on shared GPU or cloud budgets without setting limits: pilot experiments can consume compute time

Try asking

  • “Use idea-discovery on 'efficient fine-tuning for small language models' and give me a ranked idea report”
  • “Run the idea discovery pipeline on my topic with arxiv download: true”
  • “Start from zero on 'graph neural networks for materials' and validate the top ideas with pilots”
  • “Run idea discovery with pilot budget: 4h per idea, 20h total”
Step-by-step guide: how to use Idea Discovery PipelineInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    idea-discovery.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip idea-discovery.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip idea-discovery.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Set AUTO_PROCEED=false if you want to approve each stage, since the default true runs all phases without pausing
  • Review the pilot budget constants (PILOT_MAX_HOURS, MAX_TOTAL_GPU_HOURS) before starting on paid GPU resources
  • Keep ARXIV_DOWNLOAD false unless you need PDFs, to reduce bandwidth and storage

What's inside

  • SKILL.md

Redistributed unmodified from wanshuiyin/Auto-claude-code-research-in-sleep under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.