Paper AI Search
Use when the user describes a research need in natural language and wants AI-assisted query expansion before searching for papers, including requests to find recent work on a topic, related papers in a direction, or similar Chinese requests for AI paper search.
Best use case
Paper AI Search is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Use when the user describes a research need in natural language and wants AI-assisted query expansion before searching for papers, including requests to find recent work on a topic, related papers in a direction, or similar Chinese requests for AI paper search.
Teams using Paper AI Search should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/paper-ai-search/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How Paper AI Search Compares
| Feature / Agent | Paper AI Search | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Use when the user describes a research need in natural language and wants AI-assisted query expansion before searching for papers, including requests to find recent work on a topic, related papers in a direction, or similar Chinese requests for AI paper search.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Paper AI Search Use this skill when the user gives a research question, a direction, or a vague paper-finding intent rather than a ready-made keyword list. The core workflow is query expansion first, then cross-source retrieval. ## Workflow 1. Rewrite the user's natural-language request into a cleaner academic search intent. 2. Extract `3-6` compact keywords, each usually `1-3` words, favoring terms that would plausibly appear in titles or abstracts. 3. Produce one refined semantic query sentence that captures the intent at a higher level. 4. Search across all relevant academic sources with the expanded keywords rather than staying on a single source. 5. Show the expanded keywords back to the user so the search process is inspectable and easy to refine. ## Query Expansion Rules - Cover the central technical concepts rather than relying on generic words. - Mix narrower technical terms with slightly broader related terms to balance precision and recall. - Avoid low-information tokens such as `method`, `model`, or `improvement` unless they are part of a recognized phrase. - If the request is underspecified, infer and separate the task, method family, and application setting. ## Retrieval Sources Search these sources by default: - `arxiv` - `huggingface` - `semantic-scholar` - `biorxiv` - `pubmed` - `pubchem` ## Output Structure - Briefly restate the actual search intent. - List the `Expanded keywords` explicitly. - Include a `Semantic query` line when it adds clarity. - Present the ranked paper shortlist. Each entry should include: - original English title - authors - date - source - one sentence on why it matches the user's intent - End with `Next step` guidance: - narrow the scope - branch into a promising subtopic - or move straight to summary / deep reading ## Quality Rules - Do not present query expansion as ground truth. It is only a retrieval strategy. - Do not pad the answer with papers that are visibly unrelated to the expanded query. - If the results are diffuse, state that the user request is still too broad. - Preserve technical terms in their standard English form when that improves precision.
Related Skills
substructure_activity_search
Substructure-Activity Relationship - Analyze substructure-activity: ChEMBL substructure search, activity data, PubChem compounds, and similarity. Use this skill for medicinal chemistry tasks involving get substructure by smiles search activity search pubchem by smiles calculate smiles similarity. Combines 4 tools from 3 SCP server(s).
scientific-literature-search
Search scientific literature and research papers using FlowSearch to find relevant academic articles and publications.
Researcher Rigor Gate
Use before plan submission, major plan revision, and major stage transitions. Verify alignment, feasibility, rigor, completeness, and prevent unjustified regressions to earlier workflow phases.
Researcher Replan And Recovery
Use when the workflow hits contradictions, missing evidence, failed runs, design flaws, or resource shifts. Diagnose the failure class, choose the narrowest safe correction, and escalate to the user when the core plan changes.
Researcher Plan Architect
Use when the Researcher must convert a confirmed scientific goal into a staged, executable research plan with role assignments, milestones, resources, checkpoints, and risk controls.
Researcher Dispatch Supervisor
Use after the user confirms the plan. Dispatch the next justified worker task, supervise progress, enforce artifact-backed completion, and keep the workflow aligned with the approved plan.
Researcher Context Audit
Use when the Researcher starts, resumes, or reaches a major decision point. Build a context inventory from workstation materials, prior messages, existing artifacts, requirements, and unfinished work.
Researcher Ambiguity Gate
Use when the research goal, evaluation target, scope, resources, timeline, or decision criteria are ambiguous, conflicting, or not operationally testable.
Research Ideation Full
Use when the user wants the full research ideation workflow grounded in one seed paper, including complete ideation, feasibility review, experiment planning, and final synthesis, or makes an equivalent ideation request in another language.
pubmed-article-search
Search PubMed database for scientific articles and publications to retrieve biomedical literature.
pubchem-smiles-search
Search PubChem database using SMILES strings to retrieve compound information and chemical properties.
protein_similarity_search
Protein Similarity Search - Search for similar proteins: extract sequence from PDB, search structures with FoldSeek, find homologs with STRING, and check UniProt. Use this skill for bioinformatics tasks involving extract pdb sequence foldseek search get best similarity hits between species search uniprotkb entries. Combines 4 tools from 3 SCP server(s).