Pharma Market Intel Agent - FAERS Query Skill

Overview

Query real-world post-market safety data for drugs. Useful for market intel on safety profiles, emerging risks, competitor analysis.

Key outputs:

• JSON summaries (trends, top reactions/outcomes)
• PNG bar charts (yearly reports, top 10 reactions/outcomes)
• Sample recent events

Rate limits: openFDA ~240 req/min. Counts are fast (no full data).

Chemistry-Query Structure

Parse user queries into this model for standardized chaining:

from dataclasses import dataclass
from typing import List, Optional

@dataclass
class ChemistryQuery:
    drug: str  # Drug name or SMILES
    query_type: str = 'faers'  # 'faers', 'pubchem', etc.
    metrics: Optional[List[str]] = None  # ['yearly_trends', 'top_reactions', 'top_outcomes', 'events']
    limit_events: int = 20

Example:

{
  \"drug\": \"aspirin\",  // or \"CC(=O)OC1=CC=CC=C1C(=O)O\"
  \"query_type\": \"faers\",
  \"metrics\": [\"yearly_trends\", \"top_reactions\"]
}

Quick Start / Workflows

1. Basic Query (All Metrics)

exec skills/pharma-market-intel-agent/scripts/query_faers.py --drug aspirin --output ./aspirin_faers

Generates:

• aspirin_faers/aspirin_summary.json
• *.png plots
• Recent events JSON

2. SMILES Input

exec ... --drug \"CC(=O)OC1=CC=CC=C1C(=O)O\"  # Aspirin SMILES

Auto-resolves to name via PubChem.

3. Custom Limit

exec ... --drug ozempic --limit-events 50 --output ozempic_analysis

Chaining Examples

• With chemistry-query: Resolve/validate SMILES first, then FAERS.
• pharma-tox-agent: Feed top reactions for tox prediction.
• pharma-ip-expansion-agent: Check safety for IP expansion targets.
• traction-agent: Market risk scoring from FAERS trends.

# Agent workflow:
1. Parse ChemistryQuery
2. Resolve SMILES if needed (pubchempy or query_faers handles)
3. Run query_faers.py
4. Read PNGs/JSONs into response
5. Chain if metrics require

ClinicalTrials.gov Integration

Query clinical trial data from ClinicalTrials.gov API v2. Search by drug, condition, phase, and status. No API key needed.

Quick Start

# Search by drug
exec skills/pharma-market-intel-agent/scripts/query_trials.py --drug "sotorasib" --output ./sotorasib_trials

# Search by condition + filters
exec ... --condition "breast cancer" --phase PHASE3 --status RECRUITING --limit 10 --output ./bc_trials

# Search by both
exec ... --drug "pembrolizumab" --condition "NSCLC" --output ./pembro_trials

# SMILES input (auto-resolves via PubChem)
exec ... --drug "CC(=O)OC1=CC=CC=C1C(=O)O" --output ./aspirin_trials

Outputs

• {drug}_trials_summary.json — Full structured summary with trials list and aggregate stats
• {drug}_trials_by_phase.png — Bar chart by phase
• {drug}_trials_by_status.png — Bar chart by status
• {drug}_trials_timeline.png — Timeline of trial start dates