Fda 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: fda-database
description: "Query openFDA API for drugs, devices, adverse events, recalls, regulatory submissions (510k, PMA), substance identification (UNII), for FDA regulatory data analysis and safety research."
---

# FDA Database Access

## Overview

Access comprehensive FDA regulatory data through openFDA, the FDA's initiative to provide open APIs for public datasets. Query information about drugs, medical devices, foods, animal/veterinary products, and substances using Python with standardized interfaces.

**Key capabilities:**
- Query adverse events for drugs, devices, foods, and veterinary products
- Access product labeling, approvals, and regulatory submissions
- Monitor recalls and enforcement actions
- Look up National Drug Codes (NDC) and substance identifiers (UNII)
- Analyze device classifications and clearances (510k, PMA)
- Track drug shortages and supply issues
- Research chemical structures and substance relationships

## When to Use This Skill

This skill should be used when working with:
- **Drug research**: Safety profiles, adverse events, labeling, approvals, shortages
- **Medical device surveillance**: Adverse events, recalls, 510(k) clearances, PMA approvals
- **Food safety**: Recalls, allergen tracking, adverse events, dietary supplements
- **Veterinary medicine**: Animal drug adverse events by species and breed
- **Chemical/substance data**: UNII lookup, CAS number mapping, molecular structures
- **Regulatory analysis**: Approval pathways, enforcement actions, compliance tracking
- **Pharmacovigilance**: Post-market surveillance, safety signal detection
- **Scientific research**: Drug interactions, comparative safety, epidemiological studies

## Quick Start

### 1. Basic Setup

```python
from scripts.fda_query import FDAQuery

# Initialize (API key optional but recommended)
fda = FDAQuery(api_key="YOUR_API_KEY")

# Query drug adverse events
events = fda.query_drug_events("aspirin", limit=100)

# Get drug labeling
label = fda.query_drug_label("Lipitor", brand=True)

# Search device recalls
recalls = fda.query("device", "enforcement",
                   search="classification:Class+I",
                   limit=50)
```

### 2. API Key Setup

While the API works without a key, registering provides higher rate limits:
- **Without key**: 240 requests/min, 1,000/day
- **With key**: 240 requests/min, 120,000/day

Register at: https://open.fda.gov/apis/authentication/

Set as environment variable:
```bash
export FDA_API_KEY="your_key_here"
```

### 3. Running Examples

```bash
# Run comprehensive examples
python scripts/fda_examples.py

# This demonstrates:
# - Drug safety profiles
# - Device surveillance
# - Food recall monitoring
# - Substance lookup
# - Comparative drug analysis
# - Veterinary drug analysis
```

## FDA Database Categories

### Drugs

Access 6 drug-related endpoints covering the full drug lifecycle from approval to post-market surveillance.

**Endpoints:**
1. **Adverse Events** - Reports of side effects, errors, and therapeutic failures
2. **Product Labeling** - Prescribing information, warnings, indications
3. **NDC Directory** - National Drug Code product information
4. **Enforcement Reports** - Drug recalls and safety actions
5. **Drugs@FDA** - Historical approval data since 1939
6. **Drug Shortages** - Current and resolved supply issues

**Common use cases:**
```python
# Safety signal detection
fda.count_by_field("drug", "event",
                  search="patient.drug.medicinalproduct:metformin",
                  field="patient.reaction.reactionmeddrapt")

# Get prescribing information
label = fda.query_drug_label("Keytruda", brand=True)

# Check for recalls
recalls = fda.query_drug_recalls(drug_name="metformin")

# Monitor shortages
shortages = fda.query("drug", "drugshortages",
                     search="status:Currently+in+Shortage")
```

**Reference:** See `references/drugs.md` for detailed documentation

### Devices

Access 9 device-related endpoints covering medical device safety, approvals, and registrations.

**Endpoints:**
1. **Adverse Events** - Device malfunctions, injuries, deaths
2. **510(k) Clearances** - Premarket notifications
3. **Classification** - Device categories and risk classes
4. **Enforcement Reports** - Device recalls
5. **Recalls** - Detailed recall information
6. **PMA** - Premarket approval data for Class III devices
7. **Registrations & Listings** - Manufacturing facility data
8. **UDI** - Unique Device Identification database
9. **COVID-19 Serology** - Antibody test performance data

**Common use cases:**
```python
# Monitor device safety
events = fda.query_device_events("pacemaker", limit=100)

# Look up device classification
classification = fda.query_device_classification("DQY")

# Find 510(k) clearances
clearances = fda.query_device_510k(applicant="Medtronic")

# Search by UDI
device_info = fda.query("device", "udi",
                       search="identifiers.id:00884838003019")
```

**Reference:** See `references/devices.md` for detailed documentation

### Foods

Access 2 food-related endpoints for safety monitoring and recalls.

**Endpoints:**
1. **Adverse Events** - Food, dietary supplement, and cosmetic events
2. **Enforcement Reports** - Food product recalls

**Common use cases:**
```python
# Monitor allergen recalls
recalls = fda.query_food_recalls(reason="undeclared peanut")

# Track dietary supplement events
events = fda.query_food_events(
    industry="Dietary Supplements")

# Find contamination recalls
listeria = fda.query_food_recalls(
    reason="listeria",
    classification="I")
```

**Reference:** See `references/foods.md` for detailed documentation

### Animal & Veterinary

Access veterinary drug adverse event data with species-specific information.

**Endpoint:**
1. **Adverse Events** - Animal drug side effects by species, breed, and product

**Common use cases:**
```python
# Species-specific events
dog_events = fda.query_animal_events(
    species="Dog",
    drug_name="flea collar")

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 (1)

return hashlib.md5(cache_string.encode()).hexdigest()

Gates applied: no_behavioural_pass.