aminer-open-skill

English | 中文

An AMiner Skill collection for finding papers, building reading lists and taxonomies, tracing sources, and checking citations in Claude Code, Codex, OpenClaw, and similar AI assistants.

• 🧰 Choose a Skill
• 🚀 Quick Start
• 💬 Use Cases
• ℹ️ Notes
• 📚 References

Choose a Skill

A typical workflow moves from finding literature, to understanding a paper's sources, to verifying its citations. Start with the category that matches your current task.

🔎 Find and collect literature

Skill	Use it when you want to	Token	Guide
aminer-free-academic	Find and screen papers, scholars, institutions, venues, or patents with free AMiner APIs	A token is still required; listed APIs are free	SKILL.md
aminer-academic-search	Run full academic searches, deeper analysis, or explicit structured experiment retrieval across papers, scholars, institutions, venues, and patents	Required; some APIs are billed; experiment search is ¥0.10/call	SKILL.md
aminer-daily-paper	Get personalized paper recommendations from topics, scholars, authors, or an AMiner account	Required	SKILL.md
aminer-deep-search	Build a large survey bibliography through multi-round search, deduplication, and citation expansion	Required	SKILL.md
deep-research	Produce a cited, hierarchically numbered research report plus a reusable Evidence Ledger from AMiner data and the host's web tools	Required; free-first, billed APIs capped (¥20 hard stop)	SKILL.md

🗂️ Organize paper collections

Skill	Use it when you want to	Token	Guide
scientific-taxonomy	Build a hierarchical Markdown taxonomy from supplied paper titles and abstracts	Optional; only needed when AMiner must fill missing metadata	SKILL.md

🧭 Analyze and trace a paper

Skill	Use it when you want to	Token	Guide
paper-source-trace	Trace a paper's key claims to its citation contexts and sources, then generate evidence reports and citation maps	Optional; only needed for AMiner enrichment	SKILL.md / Usage guide
aminer-pdf-ocr	Convert a paper PDF to Markdown via the AMiner PDF OCR service, then have the Agent extract experiment methods, datasets, and metrics with the built-in prompt	Required	SKILL.md
aminer-exp-extraction	Extract structured experiment JSON (methods, datasets, metrics, key results) from paper Markdown via a deterministic two-stage GLM pipeline (sentence filter + extraction); if you only have a PDF, run aminer-pdf-ocr first to get the Markdown	Requires BIGMODEL_API_KEY (Zhipu BigModel); no AMiner token needed	SKILL.md

✅ Verify citations

Skill	Use it when you want to	Token	Guide
pdf-citation-verifier	Check whether the references listed in a paper PDF actually exist	Required	SKILL.md
structured-reference-audit	Build a GROBID-backed reference ledger, then conservatively resolve each parsed entry through AMiner	Required for AMiner resolution; GROBID or TEI required	SKILL.md
citation-faithfulness	Check whether in-text citations accurately represent what the cited sources say	No AMiner token required; web access is required	SKILL.md

Quick Start

1. 📦 Add the Skill you need

Clone the repository:

git clone https://github.com/CanXiangCC/aminer-open-skill.git
cd aminer-open-skill

Add the selected skills/<skill-name>/ directory to your AI assistant using its normal Skill or plugin installation method. If your assistant supports Claude plugins, use .claude-plugin/marketplace.json as the plugin list.

2. 🔑 Configure the token when required

Generate a token in the AMiner Console, then set AMINER_API_KEY in the environment used by your assistant:

export AMINER_API_KEY="<YOUR_TOKEN>"

For local Claude Code, Codex, and similar conversational sessions, the repository provides quick setup tools:

Windows:       tools\setup-aminer-token.cmd
PowerShell:    .\tools\setup-aminer-token.ps1
macOS/Linux:   ./tools/setup-aminer-token.sh

OpenClaw uses its own environment configuration:

openclaw config set env.vars.AMINER_API_KEY "<YOUR_TOKEN>"

Standalone CLI commands, CI jobs, and scheduled tasks must set the token where the command runs. Check the table above first because not every Skill requires a token.

3. 💬 Ask naturally

Describe the academic task and provide any required paper, topic, scholar, or output preferences:

Find recent papers on multimodal agents and summarize the main research directions.

Skills that provide slash commands can also be invoked directly:

/aminer-deep-search topic: "multimodal agents" target-size: 200

Use Cases

The following prompts can be used directly after the corresponding Skill is installed:

Goal	Example request	Skill
Find papers quickly	"Find 10 recent papers on long-context language models. Return the title, year, venue, citation count, and URL."	aminer-free-academic
Investigate a scholar or research topic	"Build a research profile for Andrew Ng, including interests, representative papers, collaborators, and recent work."	aminer-academic-search
Retrieve structured experiments	"Return the original Experiment JSON for this paper, filtered by method and dataset name."	aminer-academic-search
Get a focused reading list	"Recommend 8 papers on tool-using multimodal agents, prioritizing recent and highly cited work."	aminer-daily-paper
Collect literature for a survey	"Collect 200 candidate papers on retrieval-augmented generation, expand from strong seed papers, remove duplicates, and export the bibliography."	aminer-deep-search
Organize a paper collection	"Group these paper titles and abstracts into a coherent hierarchical taxonomy and save it as taxonomy.md."	scientific-taxonomy
Produce a sourced research report	"Research the Chinese LLM industry and write a cited report with an evidence ledger backing every claim."	deep-research
Trace a paper's sources	"Analyze this PDF. Trace each key claim to its citation context and source, explain the citation intent, and generate the Markdown, JSON, SVG, and HTML outputs."	paper-source-trace
Parse a PDF and extract experiments	"OCR this paper PDF, convert it to Markdown, and extract the experiment methods, datasets, and metrics as JSON."	aminer-pdf-ocr
Extract experiments from Markdown	"Read this paper's Markdown file and extract the experiment methods, datasets, metrics, and key results as one JSON file."	aminer-exp-extraction
Detect fabricated references	"Check every reference in this paper PDF and flag entries that are missing, suspicious, or need manual review."	pdf-citation-verifier
Audit parsed bibliography entries conservatively	"Parse this PDF into a reference ledger, then resolve each eligible reference through AMiner without treating an unresolved match as fabrication."	structured-reference-audit
Check citation faithfulness	"For each in-text citation in this paper, retrieve the cited source and determine whether the surrounding claim is supported by the original text."	citation-faithfulness

Notes

• Never print, log, or commit AMINER_API_KEY.
• aminer-exp-extraction calls the Zhipu BigModel API instead: set BIGMODEL_API_KEY for it (OPENAI_API_KEY accepted as fallback). It never uses the AMiner token.
• Free AMiner APIs still require a token. Start with aminer-free-academic before using paid APIs.
• Some tasks may incur API charges. Review the estimated cost before approving large searches or expensive calls.
• The setup tools under tools/ configure local chat sessions only. Set the token separately for OpenClaw, CLI, CI, and scheduled jobs.
• For direct API integration rather than Skill usage, see the AMiner Open Platform documentation below.

References

• AMiner Console
• AMiner Open Platform Documentation
• Claude plugin list
• MIT License