Openalex DatabaseSAFE

Kosmos: An AI Scientist for Autonomous Discovery - An implementation and adaptation to be driven by Claude Code or API - Based on the Kosmos AI Paper - https://arxiv.org/abs/2511.02824

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: openalex-database
description: Query and analyze scholarly literature using the OpenAlex database. This skill should be used when searching for academic papers, analyzing research trends, finding works by authors or institutions, tracking citations, discovering open access publications, or conducting bibliometric analysis across 240M+ scholarly works. Use for literature searches, research output analysis, citation analysis, and academic database queries.
---

# OpenAlex Database

## Overview

OpenAlex is a comprehensive open catalog of 240M+ scholarly works, authors, institutions, topics, sources, publishers, and funders. This skill provides tools and workflows for querying the OpenAlex API to search literature, analyze research output, track citations, and conduct bibliometric studies.

## Quick Start

### Basic Setup

Always initialize the client with an email address to access the polite pool (10x rate limit boost):

```python
from scripts.openalex_client import OpenAlexClient

client = OpenAlexClient(email="[email protected]")
```

### Installation Requirements

Install required package using uv:

```bash
uv pip install requests
```

No API key required - OpenAlex is completely open.

## Core Capabilities

### 1. Search for Papers

**Use for**: Finding papers by title, abstract, or topic

```python
# Simple search
results = client.search_works(
    search="machine learning",
    per_page=100
)

# Search with filters
results = client.search_works(
    search="CRISPR gene editing",
    filter_params={
        "publication_year": ">2020",
        "is_oa": "true"
    },
    sort="cited_by_count:desc"
)
```

### 2. Find Works by Author

**Use for**: Getting all publications by a specific researcher

Use the two-step pattern (entity name → ID → works):

```python
from scripts.query_helpers import find_author_works

works = find_author_works(
    author_name="Jennifer Doudna",
    client=client,
    limit=100
)
```

**Manual two-step approach**:
```python
# Step 1: Get author ID
author_response = client._make_request(
    '/authors',
    params={'search': 'Jennifer Doudna', 'per-page': 1}
)
author_id = author_response['results'][0]['id'].split('/')[-1]

# Step 2: Get works
works = client.search_works(
    filter_params={"authorships.author.id": author_id}
)
```

### 3. Find Works from Institution

**Use for**: Analyzing research output from universities or organizations

```python
from scripts.query_helpers import find_institution_works

works = find_institution_works(
    institution_name="Stanford University",
    client=client,
    limit=200
)
```

### 4. Highly Cited Papers

**Use for**: Finding influential papers in a field

```python
from scripts.query_helpers import find_highly_cited_recent_papers

papers = find_highly_cited_recent_papers(
    topic="quantum computing",
    years=">2020",
    client=client,
    limit=100
)
```

### 5. Open Access Papers

**Use for**: Finding freely available research

```python
from scripts.query_helpers import get_open_access_papers

papers = get_open_access_papers(
    search_term="climate change",
    client=client,
    oa_status="any",  # or "gold", "green", "hybrid", "bronze"
    limit=200
)
```

### 6. Publication Trends Analysis

**Use for**: Tracking research output over time

```python
from scripts.query_helpers import get_publication_trends

trends = get_publication_trends(
    search_term="artificial intelligence",
    filter_params={"is_oa": "true"},
    client=client
)

# Sort and display
for trend in sorted(trends, key=lambda x: x['key'])[-10:]:
    print(f"{trend['key']}: {trend['count']} publications")
```

### 7. Research Output Analysis

**Use for**: Comprehensive analysis of author or institution research

```python
from scripts.query_helpers import analyze_research_output

analysis = analyze_research_output(
    entity_type='institution',  # or 'author'
    entity_name='MIT',
    client=client,
    years='>2020'
)

print(f"Total works: {analysis['total_works']}")
print(f"Open access: {analysis['open_access_percentage']}%")
print(f"Top topics: {analysis['top_topics'][:5]}")
```

### 8. Batch Lookups

**Use for**: Getting information for multiple DOIs, ORCIDs, or IDs efficiently

```python
dois = [
    "https://doi.org/10.1038/s41586-021-03819-2",
    "https://doi.org/10.1126/science.abc1234",
    # ... up to 50 DOIs
]

works = client.batch_lookup(
    entity_type='works',
    ids=dois,
    id_field='doi'
)
```

### 9. Random Sampling

**Use for**: Getting representative samples for analysis

```python
# Small sample
works = client.sample_works(
    sample_size=100,
    seed=42,  # For reproducibility
    filter_params={"publication_year": "2023"}
)

# Large sample (>10k) - automatically handles multiple requests
works = client.sample_works(
    sample_size=25000,
    seed=42,
    filter_params={"is_oa": "true"}
)
```

### 10. Citation Analysis

**Use for**: Finding papers that cite a specific work

```python
# Get the work
work = client.get_entity('works', 'https://doi.org/10.1038/s41586-021-03819-2')

# Get citing papers using cited_by_api_url
import requests
citing_response = requests.get(
    work['cited_by_api_url'],
    params={'mailto': client.email, 'per-page': 200}
)
citing_works = citing_response.json()['results']
```

### 11. Topic and Subject Analysis

**Use for**: Understanding research focus areas

```python
# Get top topics for an institution
topics = client.group_by(
    entity_type='works',
    group_field='topics.id',
    filter_params={
        "authorships.institutions.id": "I136199984",  # MIT
        "publication_year": ">2020"
    }
)

for topic in topics[:10]:
    print(f"{topic['key_display_name']}: {topic['count']} works")
```

### 12. Large-Scale Data Extraction

**Use for**: Downloading large datasets for analysis

```python
# Paginate through all results
all_papers = client.paginate_all(
    endpoint='/works',
    params={
        'search': 'synthetic biology',
        'filter': 'publicati

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

declared (1 observation(s))

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.