Idea Creator
AI4Scientist/nano-scientist
Generate and rank research ideas given a broad direction. An agent skill from AI4Scientist/nano-scientist.
Turns a vague research direction into a focused, problem-anchored method plan through up to five review rounds with a second model.
Install Research Refine
$ npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install zjYao36/Auto-Research-Refine research-refine --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/zjYao36/Auto-Research-Refine.git skills-src && mkdir -p .claude/skills && cp -r skills-src/research-refine .claude/skills/research-refine && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "research-refine" agent skill from https://github.com/zjYao36/Auto-Research-Refine/tree/main/research-refine into .claude/skills/research-refine/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "research-refine", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/zjYao36/Auto-Research-Refine/tree/main/research-refineType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install zjYao36/Auto-Research-Refine research-refine --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/zjYao36/Auto-Research-Refine.git skills-src && mkdir -p .agents/skills && cp -r skills-src/research-refine .agents/skills/research-refine && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "research-refine" agent skill from https://github.com/zjYao36/Auto-Research-Refine/tree/main/research-refine into .agents/skills/research-refine/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "research-refine", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install zjYao36/Auto-Research-Refine research-refine --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/zjYao36/Auto-Research-Refine.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/research-refine .cursor/skills/research-refine && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "research-refine" agent skill from https://github.com/zjYao36/Auto-Research-Refine/tree/main/research-refine into .cursor/skills/research-refine/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "research-refine", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/zjYao36/Auto-Research-Refine.git --path research-refine--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install zjYao36/Auto-Research-Refine research-refine --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/zjYao36/Auto-Research-Refine.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/research-refine .gemini/skills/research-refine && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "research-refine" agent skill from https://github.com/zjYao36/Auto-Research-Refine/tree/main/research-refine into .gemini/skills/research-refine/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "research-refine", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install zjYao36/Auto-Research-Refine research-refineInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/zjYao36/Auto-Research-Refine.git skills-src && mkdir -p .github/skills && cp -r skills-src/research-refine .github/skills/research-refine && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "research-refine" agent skill from https://github.com/zjYao36/Auto-Research-Refine/tree/main/research-refine into .github/skills/research-refine/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "research-refine", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install zjYao36/Auto-Research-Refine research-refine --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/zjYao36/Auto-Research-Refine.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/research-refine .opencode/skills/research-refine && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "research-refine" agent skill from https://github.com/zjYao36/Auto-Research-Refine/tree/main/research-refine into .opencode/skills/research-refine/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "research-refine", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Facts
- Skill name
research-refine- GitHub stars
- 128
- Used in
- 6 other repos
- Token cost
- ~6.9k tokens
- SKILL.md length
- 1,776 words
- Files
- 2
- Skills in repo
- 3
- Repo updated
- First seen
- Licence
- None found
At a glance
Turns a vague research direction into a focused, problem-anchored method plan through up to five review rounds with a second model.
- Works in 6 steps: Freeze the Problem Anchor → Build the Initial Proposal → External Method Review (Round 1) → …
- Refining a fuzzy research approach into a concrete method plan
- SKILL.md covers Overview, Constants, Output Structure and Workflow, plus 2 more sections
- Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md
What it does
research-refine takes a research problem with a fuzzy approach and works it into a problem, focused method, minimal validation document. A fixed Problem Anchor is written first and reused every round so the original problem is not lost. Four principles guide it: the smallest adequate mechanism wins, one paper carries one dominant contribution, and modern techniques such as LLMs, VLMs, diffusion, RL, distillation or inference-time scaling are used only where they naturally fit the bottleneck.
Claude drafts a proposal after scanning grounding papers and identifying the technical gap, then a reviewer model, gpt-5.4 reached through Codex MCP, scores it, and the proposal is revised for up to 5 rounds or until the overall score reaches 9. Defaults cap local papers scanned at 15, core experiments at 3, primary claims at 2 and new trainable components at 2, and arguments can override them. Round files, reviews and the final report go to a refine-logs/ folder, and each refinement file must hold a full anchored proposal, not only incremental fixes.
When your agent uses it
- Refining a fuzzy research approach into a concrete method plan
- Decomposing a research problem without losing the original problem statement
- Trimming an overbuilt idea down to one dominant contribution
- Having a reviewer model score and challenge a proposal over several rounds
Example prompts
- “Refine my approach: the problem is label noise in speech data, and my idea is contrastive pretraining.”
- “Decompose this problem and give me a minimal-validation plan, with at most 3 review rounds.”
- “帮我细化方案：长视频理解里的记忆压缩问题。”
Requirements
- Codex MCP configured so the gpt-5.4 reviewer can be called
- Pre-approved tools (allowed-tools): Bash(*), Read, Write, Edit, Grep, Glob, WebSearch, WebFetch, Agent, mcp__codex__codex, mcp__codex__codex-reply
Workflow steps
6 steps, taken from the step headings in SKILL.md.
- Freeze the Problem Anchor
- Build the Initial Proposal
- External Method Review (Round 1)
- Parse Feedback and Revise the Method
- Re-evaluation (Round 2+)
- Final Report and Logs
What it can do on your machine
Read from SKILL.md and the folder at commit a1f1449. It shows what the files ask for, not the result of running them.
Tool permissions
Pre-approves these tools, so the agent can use them without asking each time:
Bash(*)ReadWriteEditGrepGlobWebSearchWebFetchAgentmcp__codex__codex
…and 1 more on the same allowed-tools line.
From allowed-tools in the SKILL.md frontmatter.
Runs code
No scripts in the folder and no shell commands in SKILL.md (its code samples are markdown).
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Context cost
Research Refine loads about 6.9k tokens when it runs. Until then it costs about 103 tokens; SKILL.md has 1,776 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
Safety
Auto-check: notes
The automated check noted patterns worth knowing about, such as sudo or a known installer.
- NotePre-approves every shell command (allowed-tools: Bash)SKILL.md
allowed-tools: Bash(*), Read, Write, Edit, Grep, Glob, WebSearch, WebFetch, Agent, mcp__codex__codex, mcp__codex__c
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
SKILL.md
Without a licence we can't republish the file, so here is its outline and opening line. It has 1,776 words (~6,934 tokens).
- name
- research-refine
- allowed-tools
- Bash(*), Read, Write, Edit, Grep, Glob, WebSearch, WebFetch, Agent, mcp__codex__codex, mcp__codex__codex-reply
Files
SKILL.md and 1 other file in research-refine of zjYao36/Auto-Research-Refine.
- SKILL.md
- agents/openai.yaml
Open the folder on GitHubat commit a1f1449
Used in 6 other repositories
We found 17 copies of this SKILL.md (exact, near-identical or edited) in other folders, from 6 other GitHub owners. This page covers the copy in zjYao36/Auto-Research-Refine, which our catalogue first saw on October 7, 2026.
- GRIND-Lab-Core/night_owl_research_agentskills/refine-research
- AI4Scientist/nano-scientistskills/research-refineskills/skills-codex/research-refine+2 more
- OpenLAIR/dr-clawskills/aris-research-refine
- wanshuiyin/Auto-claude-code-research-in-sleepskills/research-refineskills/skills-codex/research-refine+2 more
- brycewang-stanford/Auto-Empirical-Research-Skillsskills/42-wanshuiyin-ARIS/skills/research-refineskills/42-wanshuiyin-ARIS/skills/skills-codex/research-refine+2 more
- Tai609/NebulaMatruntime/skills-bundle/aris-research-refineruntime/aris/skills-nebulamat/aris-research-refine+1 more
Compare with similar skills
Research Refine next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Research Refine this skillzjYao36/Auto-Research-Refine | 128 | 6 repos | ~6.9k | Automated safety check: Notes | None | |
| Idea CreatorAI4Scientist/nano-scientist | 128 | 4 repos | ~3.9k | Automated safety check: Warn | None | |
| Idea Discovery PipelineGRIND-Lab-Core/night_owl_research_agent | 106 | — | ~4.4k | Automated safety check: Warn | None | |
| Research ReviewGRIND-Lab-Core/night_owl_research_agent | 106 | 5 repos | ~1.1k | Automated safety check: Notes | None | |
| Research Reviewwanshuiyin/Auto-claude-code-research-in-sleep | 17k | — | ~3.1k | Automated safety check: Notes | MIT | |
| Novelty CheckAI4Scientist/nano-scientist | 128 | 4 repos | ~823 | Automated safety check: Pass | None |
Similar skills
Idea Discovery Pipeline
GRIND-Lab-Core/night_owl_research_agent
The full pipeline for idea generation. An agent skill from GRIND-Lab-Core/night_owl_research_agent.
Research Review
GRIND-Lab-Core/night_owl_research_agent
Get a deep critical review of research idea from GPT via Codex MCP.
Research Review
wanshuiyin/Auto-claude-code-research-in-sleep
Get a deep critical review of research from an external reviewer backend (Codex or manual).
Novelty Check
AI4Scientist/nano-scientist
Verify research idea novelty against recent literature. An agent skill from AI4Scientist/nano-scientist.
Research Coordinator
brycewang-stanford/Auto-Empirical-Research-Skills
Research Coordinator v12.0 - Human-Centered Edition (Systematic Review Automation) Context-persistent platform with 24 specialized agents across 9 categories (A-G, I, X).
More from zjYao36/Auto-Research-Refine
Claim-Driven Experiment Planner
zjYao36/Auto-Research-Refine
Turns a refined research proposal into a claim-to-evidence-to-run-order roadmap instead of a sprawling benchmark wishlist.
Research Refine Pipeline
zjYao36/Auto-Research-Refine
Chains research-refine and experiment-plan to turn a vague research direction into a focused proposal and a claim-driven experiment roadmap.
Related
Works with
Categories
Questions about Research Refine
What does Research Refine do?
Turns a vague research direction into a focused, problem-anchored method plan through up to five review rounds with a second model. research-refine takes a research problem with a fuzzy approach and works it into a problem, focused method, minimal validation document. A fixed Problem Anchor is written first and reused every round so the original problem is not lost.
When should I use Research Refine?
Research Refine fits situations like: refining a fuzzy research approach into a concrete method plan; decomposing a research problem without losing the original problem statement; trimming an overbuilt idea down to one dominant contribution; having a reviewer model score and challenge a proposal over several rounds.
How do I install Research Refine in Claude Code?
Run `npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a claude-code`. Or copy the skill folder (research-refine in zjYao36/Auto-Research-Refine) into .claude/skills/research-refine in your project. Claude Code loads it when a task matches its description.
How do I install Research Refine in Codex?
Run `npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a codex`. Or copy the skill folder (research-refine in zjYao36/Auto-Research-Refine) into .agents/skills/research-refine in your project. Codex loads it when a task matches its description.
Can I use Research Refine in Cursor, Gemini CLI or GitHub Copilot?
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add zjYao36/Auto-Research-Refine --skill research-refine -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/research-refine, .gemini/skills/research-refine, .github/skills/research-refine and .opencode/skills/research-refine in your project.
What does Research Refine need to run?
SKILL.md names no scripts, command-line tools or credentials: Research Refine is instructions for the agent only. Our summary lists: Codex MCP configured so the gpt-5.4 reviewer can be called. Its frontmatter pre-approves these tools: Bash(*), Read, Write, Edit, Grep, Glob, WebSearch, WebFetch, Agent, mcp__codex__codex, mcp__codex__codex-reply.
Does Research Refine access the network?
SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.
Is Research Refine safe to install?
Our automated static check of SKILL.md found notes only (pre-approves every shell command (allowed-tools: bash)), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.
What licence does Research Refine use?
No licence was found for Research Refine or its repository. Without one, default copyright applies: ask the author before reusing or redistributing it.
How many tokens does Research Refine use?
About 6.9k tokens (SKILL.md is roughly 28k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
What are the alternatives to Research Refine?
Skills that share tags, products or a category with Research Refine: Idea Creator (AI4Scientist/nano-scientist, 128 stars), Idea Discovery Pipeline (GRIND-Lab-Core/night_owl_research_agent, 106 stars), Research Review (GRIND-Lab-Core/night_owl_research_agent, 106 stars) and Research Review (wanshuiyin/Auto-claude-code-research-in-sleep, 17k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Who maintains Research Refine?
zjYao36 (a GitHub user) maintains it in zjYao36/Auto-Research-Refine, which has 128 GitHub stars. The repository holds 3 skills in this directory. The repository was last updated on March 16, 2026.
Source: zjYao36/Auto-Research-Refine on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.