Agent skill

Sciatlas Idea Generate

by zjunlp in zjunlp/SciAtlas

Use only the current SciAtlas multi-step idea-generation workflow (sciatlasideagen) to take a novice user from zero setup to final literature-grounded research idea seeds, including setup…

MITAuto-check: notesAgent Workflows

Install Sciatlas Idea Generate

skills CLI
$ npx skills add zjunlp/SciAtlas --skill sciatlas-idea-generate -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install zjunlp/SciAtlas sciatlas-idea-generate --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/zjunlp/SciAtlas.git skills-src && mkdir -p .claude/skills && cp -r skills-src/agent-skill/sciatlas-idea-generate .claude/skills/sciatlas-idea-generate && rm -rf skills-src

Use ~/.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/

Facts

Skill name
sciatlas-idea-generate
GitHub stars
160
Token cost
~1.7k tokens
SKILL.md length
705 words
Files
2
Skills in repo
7
Repo updated
First seen
Licence
MIT

At a glance

Use only the current SciAtlas multi-step idea-generation workflow (sciatlasideagen) to take a novice user from zero setup to final literature-grounded research idea seeds, including setup…

  • Works in 6 steps: Check for the repository workflow → If missing, clone the full SciAtlas… → Check current environment and .env for… → …
  • The user wants new research directions
  • SKILL.md covers Operating Contract, KG Linkage, Zero-Start Bootstrap and Run Plan, plus 2 more sections
  • Calls python; reaches sciatlas.openkg.cn; needs SCIATLAS_API_KEY and LLM_API_KEY

What it does

Sciatlas Idea Generate is an agent skill from zjunlp/SciAtlas. Use only the current SciAtlas multi-step idea-generation workflow (sciatlasideagen) to take a novice user from zero setup to final literature-grounded research idea seeds, including setup, registration guidance, workflow configuration, KG retrieval, artifact reading, novelty checking, and synthesis. Trigger when the user wants new research directions, hypotheses, cross-topic combinations, project ideas, or brainstorming grounded in SciAtlas KG retrieval.

Its SKILL.md is about 1.7k tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files (for example `agents/openai.yaml`).

It sits in Agent Workflows, covering Brainstorming and Hypothesis generation. It works with Python. The repository describes itself as: A Large-Scale Knowledge Graph for Automated Scientific Research. The licence is MIT.

When your agent uses it

  • The user wants new research directions
  • Cross-topic combinations
  • Brainstorming grounded in SciAtlas KG retrieval

Example prompts

  • “/sciatlas-idea-generate”

Requirements

  • Python 3
  • A credential in SCIATLAS_API_KEY
  • A credential in LLM_API_KEY

Workflow steps

6 steps, taken from the first numbered list in SKILL.md.

  1. Check for the repository workflow
  2. If missing, clone the full SciAtlas repository, change into it, then run python -m pip install -e ./sciatlas and python -m pip install -r…
  3. Check current environment and .env for hosted KG and LLM settings before asking the user.
  4. If SCIATLAS_API_KEY is missing for hosted mode, guide the user to http://sciatlas.openkg.cn/register; ask for email, verification code…
  5. Configure hosted KG yourself
  6. Configure the LLM endpoint required by the multi-step workflow

What it can do on your machine

Read from SKILL.md and the folder at commit e8873a9. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • python

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Hosts in commands or code, which the agent is likely to contact:

    • sciatlas.openkg.cn

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names these keys or tokens, usually read from environment variables:

    • SCIATLAS_API_KEY
    • LLM_API_KEY

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Sciatlas Idea Generate loads about 1.7k tokens when it runs. Until then it costs about 121 tokens; SKILL.md has 705 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~121
When it runs · the whole SKILL.md, loaded when a task matches
~1.7k

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.

  • NoteMentions a .env fileSKILL.md:12
    workflow, guide registration, configure `.env` or shell variables, run retrieval/generation, inspect artifacts, and synt
  • NoteMentions a .env fileSKILL.md:37
    3. Check current environment and `.env` for hosted KG and LLM settings before asking the user.
  • NoteMentions a .env fileSKILL.md:61
    then write them to the current shell or `.env` without echoing secrets.
  • NoteMentions a .env fileSKILL.md:70
    ne also needs its own `references/search/.env` with Neo4j/model settings.

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

The full file from zjunlp/SciAtlas at commit e8873a9, republished under its MIT licence (© zjunlp). 705 words, ~1,691 tokens.

Download SKILL.mdSave it as .claude/skills/sciatlas-idea-generate/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
sciatlas-idea-generate
description
Use only the current SciAtlas multi-step idea-generation workflow (`sciatlas_idea_gen`) to take a novice user from zero setup to final literature-grounded research idea seeds, including setup, registration guidance, workflow configuration, KG retrieval, artifact reading, novelty checking, and synthesis. Trigger when the user wants new research directions, hypotheses, cross-topic combinations, project ideas, or brainstorming grounded in SciAtlas KG retrieval.

SciAtlas Idea Generate

Use this skill to generate research ideas from SciAtlas KG evidence. Run only the repository workflow python -m sciatlas_idea_gen.main, which retrieves anchor papers, builds a research graph, searches same-field and cross-domain inspirations, generates ideas, and runs novelty checks.

Operating Contract

  • Own the end-to-end novice flow: install or locate the workflow, guide registration, configure .env or shell variables, run retrieval/generation, inspect artifacts, and synthesize final idea seeds.
  • Ask the user only for human-only values: missing topic/domain, email, verification code, SciAtlas token, LLM key/base/model, S2 key, local KG credentials when explicitly using local KG, or one necessary scope clarification.
  • Do not ask the user to run shell commands when tool access is available.
  • Never disclose the full API token.
  • Use sciatlas_idea_gen as the only idea-generation workflow.
  • If sciatlas_idea_gen is unavailable or required LLM/KG setup is missing, report the blocker and fix setup if possible; do not substitute another route.
  • Save and inspect artifacts before writing the final answer.

KG Linkage

The workflow reaches the KG through sciatlas_idea_gen.clients.SciAtlasClient:

  • Hosted mode: SCIATLAS_USE_LOCAL_KG=0; the adapter shells into this repository's run_sciatlas.py for KG-backed paper retrieval, then normalizes data.result.ranking.papers or data.result.sources.kg.papers to papers.
  • Local KG mode: SCIATLAS_USE_LOCAL_KG=1; the adapter runs references/search/run_search.py with --disable-s2, --disable-llm-ranking, --kg-top-k, and --final-top-k, then normalizes the local KG response to papers.
  • Keep hosted mode as the default unless the user says they have Neo4j and local KG models configured.

Zero-Start Bootstrap

  1. Check for the repository workflow:
bash
python -m sciatlas_idea_gen.main -h
  1. If missing, clone the full SciAtlas repository, change into it, then run python -m pip install -e ./sciatlas and python -m pip install -r requirements-workflows.txt. Do not use the GitHub #subdirectory=sciatlas package-only installation: it does not include sciatlas_idea_gen.
  2. Check current environment and .env for hosted KG and LLM settings before asking the user.
  3. If SCIATLAS_API_KEY is missing for hosted mode, guide the user to http://sciatlas.openkg.cn/register; ask for email, verification code, and returned sciatlas_xxx token only when needed.
  4. Configure hosted KG yourself:
powershell
$env:SCIATLAS_API_BASE_URL = "http://sciatlas.openkg.cn"
$env:SCIATLAS_API_KEY = "<token>"
$env:SCIATLAS_USE_LOCAL_KG = "0"
bash
export SCIATLAS_API_BASE_URL="http://sciatlas.openkg.cn"
export SCIATLAS_API_KEY="<token>"
export SCIATLAS_USE_LOCAL_KG=0
  1. Configure the LLM endpoint required by the multi-step workflow:
bash
export LLM_API_KEY="<provider-key>"
export LLM_BASE_URL="https://your-provider.example/v1"
export LLM_MODEL="<model-name>"

Ask only for missing LLM provider values, then write them to the current shell or .env without echoing secrets.

  1. Configure local KG only when requested:
bash
export SCIATLAS_USE_LOCAL_KG=1
export SCIATLAS_LOCAL_SEARCH_ROOT="./references/search"

The local KG engine also needs its own references/search/.env with Neo4j/model settings.

Show full SKILL.md (335 more words)Show less

Run Plan

Use the flash path first for quick, interactive idea generation:

bash
python -m sciatlas_idea_gen.main "<research topic>" --workflow flash --domain "<optional field>"

Run the full path when the user wants a broader research graph, more seed diversity, and a more comprehensive inspiration search:

bash
python -m sciatlas_idea_gen.main "<research topic>" --workflow full --domain "<optional field>"

For a seed PDF:

bash
python -m sciatlas_idea_gen.main "<research topic>" --workflow full --pdf path/to/paper.pdf

Useful overrides:

  • --workflow flash for a faster path: 1 seed query, a compact graph, one same-field inspiration, one cross-domain inspiration, compressed Step 5/Step 8 gate stages, and no novelty feedback retry.
  • --workflow full for the broader path: multi-query seed retrieval, larger graph construction, broader inspiration retrieval, and novelty feedback retry.
  • --k-step1 N for more seed papers.
  • --anchor-top-k N for broader Step 1 retrieval.
  • --graph-budget-min N and --graph-budget-max N for graph size.
  • --seed-recent-years N to include older foundational papers.
  • --resume-from-run-dir runs/<id> to continue from saved artifacts.

Artifacts To Read

Read the run directory before answering:

  • summary.json: status, effective config, counts, and final ideas.
  • retrieval_trace.json: KG/S2/OpenAlex retrieval events and returned papers.
  • step1_seed_papers.json: anchor papers and refined query.
  • step2_research_graph.json: graph nodes, edges, evidence scores, and phases.
  • step3_trend.txt: trend summary from the graph.
  • step5_radius_plan.json: inspiration radius plan. In flash, this is a deterministic compact plan and summary.json may mark Step 5 as compressed.
  • step6_inspiration_candidates.json, step7_inspirations.json, step8_selected_inspirations.json: inspiration path.
  • step9_ideas.json and step9_ideas.md: final ideas and citations.

In flash, treat status: "compressed" for Step 5 or Step 8 as normal success. Step 5 skips the LLM radius gate and uses the compact same-field/cross-domain plan. Step 8 skips the LLM selector and keeps the compact plausible inspiration set. Do not rerun full merely because these stages are compressed.

If the run is in smoke mode and fails partway, use the completed artifacts and clearly state the failed step. Smoke mode is now diagnostic; prefer --workflow flash for normal quick runs.

Deliverable

Return 3-8 idea seeds when available. For each idea:

  • Name.
  • One-sentence hypothesis.
  • KG evidence trail from seed papers, graph papers, and inspirations.
  • Why it is nontrivial.
  • Closest-prior-art or novelty risk.
  • Minimal validation experiment.
  • Next SciAtlas query or pipeline setting to test novelty.

Keep speculative claims clearly labeled and cite artifacts/paper titles from the run.

© zjunlp, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file in agent-skill/sciatlas-idea-generate of zjunlp/SciAtlas.

  • SKILL.md
  • agents/openai.yaml

Open the folder on GitHubat commit e8873a9

Compare with similar skills

Sciatlas Idea Generate 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.

Sciatlas Idea Generate compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Sciatlas Idea Generate this skillzjunlp/SciAtlas160—~1.7kAutomated safety check: NotesMIT
News to Research Idea BriefingOpenLAIR/dr-claw1.2k—~1.3kAutomated safety check: NotesCustom licence
Video Copy AnalyzerALBEDO-TABAI/video-copy-analyzer209—~1.9kAutomated safety check: PassMIT
Generate IdeaGRIND-Lab-Core/night_owl_research_agent106—~3.2kAutomated safety check: WarnNone
Research Ideation ScreeningBingHanOfUESTC/open_agent_team106—~492Automated safety check: PassMIT
Research Junshijunshi-research/research-junshi126—~2.1kAutomated safety check: PassApache-2.0

Similar skills

  • Clusters the latest news-feed results by topic and writes a briefing of research idea seeds with citations, plus a structured seeds file, without crawling new sources.

    1.2k GitHub stars~1.3k tokensUpdated 21 days ago
    Research & ScienceAuto-check: notes
  • Video Copy Analyzer

    ALBEDO-TABAI/video-copy-analyzer

    视频文案分析一站式工具。下载在线视频(B站/YouTube/抖音等)、使用FunASR进行高速中文语音转录、 自动校正文稿、并进行三维度综合分析(TextContent/Viral/Brainstorming)。

    209 GitHub stars~1.9k tokensUpdated 1 mo ago
    Agent WorkflowsAuto-check passed
  • Generate Idea

    GRIND-Lab-Core/night_owl_research_agent

    Generate and rank research ideas given a broad direction. An agent skill from GRIND-Lab-Core/night_owl_research_agent.

    106 GitHub stars~3.2k tokensUpdated 5 mo ago
    Agent WorkflowsAuto-check: warnings
  • Research Ideation Screening

    BingHanOfUESTC/open_agent_team

    A skill your agent uses when generating, screening, ranking, or selecting research ideas from literature gaps and experimental constraints.

    106 GitHub stars~492 tokensUpdated 3 mo ago
    Agent WorkflowsAuto-check passed
  • Research Junshi

    junshi-research/research-junshi

    Personalized academic literature digests and ranked research ideas grounded in the researcher's papers and persistent memory.

    126 GitHub stars~2.1k tokensUpdated 22 days ago
    Agent WorkflowsAuto-check passed
  • Idea Generation

    voidful/academic-skills

    學術研究的 Idea 產生技能——從發散到收斂,系統化地產出高品質研究構想。當使用者想腦力激盪研究方向、找新 research idea、或問「我接下來可以做什麼研究」時,一定要使用此技能。觸發詞包括:brainstorm、想 idea、研究方向、下一步做什麼、有什麼可以研究的、找 gap、research proposal。適用於任何階段的學術研究構想生成。

    134 GitHub stars~1.6k tokensUpdated 6 mo ago
    Agent WorkflowsAuto-check passed

More from zjunlp/SciAtlas

  • Sciatlas Idea Evaluate

    zjunlp/SciAtlas

    Use only the current SciAtlas automated review workflow (reviewpipeline) to take a novice user from zero setup to a final automated review of a research idea or paper, including setup, registration…

    160 GitHub stars~2.2k tokensUpdated 9 days ago
    Auto-check: notes
  • Use only SciAtlas search-papers to take a novice user from zero setup to final prior-art grounding for a research idea, including setup, registration guidance, configuration, retrieval, artifact…

    160 GitHub stars~1.1k tokensUpdated 9 days ago
    Auto-check passed
  • Use only the current SciAtlas literature-review workflow (literaturereviewpipeline) to take a novice user from zero setup to a final evidence-grounded survey outline, paper map, or literature…

    160 GitHub stars~1.7k tokensUpdated 9 days ago
    Auto-check: notes
  • Use only SciAtlas search-papers to take a novice user from zero setup to a final small evidence-backed paper search result, including setup, registration guidance, configuration, retrieval, artifact…

    160 GitHub stars~1.3k tokensUpdated 9 days ago
    Auto-check passed
  • Use only SciAtlas search-papers to take a novice user from zero setup to a final evidence-grounded researcher profile from retrieved papers, including setup, registration guidance, configuration…

    160 GitHub stars~1.1k tokensUpdated 9 days ago
    Auto-check passed
  • Sciatlas Trend Report

    zjunlp/SciAtlas

    Use only SciAtlas search-papers to take a novice user from zero setup to a final timeline-oriented research trend report, including setup, registration guidance, configuration, retrieval, artifact…

    160 GitHub stars~965 tokensUpdated 9 days ago
    Auto-check passed

Works with

Categories

Questions about Sciatlas Idea Generate

What does Sciatlas Idea Generate do?

Use only the current SciAtlas multi-step idea-generation workflow (sciatlasideagen) to take a novice user from zero setup to final literature-grounded research idea seeds, including setup…. Sciatlas Idea Generate is an agent skill from zjunlp/SciAtlas. Use only the current SciAtlas multi-step idea-generation workflow (sciatlasideagen) to take a novice user from zero setup to final literature-grounded research idea seeds, including setup, registration guidance, workflow configuration, KG retrieval, artifact reading, novelty checking, and synthesis.

When should I use Sciatlas Idea Generate?

Sciatlas Idea Generate fits situations like: the user wants new research directions; cross-topic combinations; brainstorming grounded in SciAtlas KG retrieval.

How do I install Sciatlas Idea Generate in Claude Code?

Run `npx skills add zjunlp/SciAtlas --skill sciatlas-idea-generate -a claude-code`. Or copy the skill folder (agent-skill/sciatlas-idea-generate in zjunlp/SciAtlas) into .claude/skills/sciatlas-idea-generate in your project. Claude Code loads it when a task matches its description.

How do I install Sciatlas Idea Generate in Codex?

Run `npx skills add zjunlp/SciAtlas --skill sciatlas-idea-generate -a codex`. Or copy the skill folder (agent-skill/sciatlas-idea-generate in zjunlp/SciAtlas) into .agents/skills/sciatlas-idea-generate in your project. Codex loads it when a task matches its description.

Can I use Sciatlas Idea Generate 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 zjunlp/SciAtlas --skill sciatlas-idea-generate -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/sciatlas-idea-generate, .gemini/skills/sciatlas-idea-generate, .github/skills/sciatlas-idea-generate and .opencode/skills/sciatlas-idea-generate in your project.

What does Sciatlas Idea Generate need to run?

Going by SKILL.md and its folder, Sciatlas Idea Generate needs the command-line tools its instructions call (python) and credentials named SCIATLAS_API_KEY and LLM_API_KEY. Our summary lists: Python 3; A credential in SCIATLAS_API_KEY; A credential in LLM_API_KEY.

Does Sciatlas Idea Generate access the network?

SKILL.md names 1 domain. In commands or code: sciatlas.openkg.cn; the agent is likely to contact it when it follows the instructions. This is read from the text; nothing was executed.

Is Sciatlas Idea Generate safe to install?

Our automated static check of SKILL.md found notes only (mentions a .env file), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.

What licence does Sciatlas Idea Generate use?

Sciatlas Idea Generate is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Sciatlas Idea Generate use?

About 1.7k tokens (SKILL.md is roughly 6.8k 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 Sciatlas Idea Generate?

Skills that share tags, products or a category with Sciatlas Idea Generate: News to Research Idea Briefing (OpenLAIR/dr-claw, 1.2k stars), Video Copy Analyzer (ALBEDO-TABAI/video-copy-analyzer, 209 stars), Generate Idea (GRIND-Lab-Core/night_owl_research_agent, 106 stars) and Research Ideation Screening (BingHanOfUESTC/open_agent_team, 106 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Sciatlas Idea Generate?

zjunlp (a GitHub organization) maintains it in zjunlp/SciAtlas, which has 160 GitHub stars. The repository holds 7 skills in this directory. The repository was last updated on September 30, 2026.

Source: zjunlp/SciAtlas on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.