Back to Strategies
Agent (Categorical)

Multi-Step Agent (Categorical)

Agent (Categorical)Claude & GPT-4

5-step agent pipeline: Extract Facts > Web Search > Clinical Assessment > Market Dynamics > Final Prediction. Uses specialized prompts at each step and aggregates findings for a more informed prediction.

Overview

Strategy Type
Agent (Categorical)
Number of Steps
5 steps
Models Used
Claude & GPT-4
Output Format
Categorical (very_negative to very_positive)

Pipeline Visualization

1

Step 1: Extract Facts

LLM prompt execution

2

Step 2: Web Search

Automated web search (no LLM call)

3

Step 3: Clinical Assessment

LLM prompt execution

4

Step 4: Market Dynamics

LLM prompt execution

5

Step 5: Final Prediction

LLM prompt execution

Prompt Details

1

Step 1: Extract Facts

You are a biotech analyst extracting key facts from press releases. Be precise and factual.
Extract the key clinical/regulatory facts from this press release. PRESS RELEASE: {pr_text} Return a JSON object with: {{ "primary_endpoint_met": true/false/null, "p_value": "<p-value if mentioned, else null>", "hazard_ratio": "<HR if mentioned, else null>", "confidence_interval": "<CI if mentioned, else null>", "effect_size": "<key efficacy metric>", "safety_signals": ["list of safety concerns mentioned"], "regulatory_status": "<FDA status or null>", "trial_phase": "<Phase 1/2/3 or null>", "patient_population": "<description of patients>", "key_quotes": ["1-2 important quotes"] }}

Expected Output Format

This prompt expects a JSON response. See the user prompt template for the exact structure.

2

Step 2: Web Search

Automated
(Automated web search step - not an LLM prompt)
Search for {ticker} stock biotech news and {drug} {company} analyst expectations
3

Step 3: Clinical Assessment

You are a clinical trial expert assessing the medical significance of trial results.
Based on these extracted facts, assess the clinical significance. EXTRACTED FACTS: {step1_result} CATALYST INFO: - Drug: {drug} - Phase: {phase} - Indication: {indication} Answer these questions in JSON: {{ "clinical_significance": "breakthrough/meaningful/incremental/neutral/disappointing/failure", "vs_standard_of_care": "<comparison>", "unmet_need_addressed": "<assessment>", "approvability_assessment": "<likely/uncertain/unlikely>", "clinical_reasoning": "<2-3 sentences>" }}

Expected Output Format

This prompt expects a JSON response. See the user prompt template for the exact structure.

4

Step 4: Market Dynamics

You are a biotech equity analyst evaluating market dynamics and investor expectations.
Evaluate the market dynamics for this catalyst. CLINICAL ASSESSMENT: {step3_result} MARKET CONTEXT FROM WEB: {step2_result} CATALYST INFO: - Ticker: {ticker} - Company: {company} - Drug: {drug} - Phase: {phase} - Event Type: {cat_type} Return JSON: {{ "priced_in_assessment": "fully_priced_in/partially_priced_in/not_priced_in/negatively_positioned", "market_expectations": "exceeded/met/missed", "asset_importance": "lead_asset/key_pipeline/minor_program", "market_opportunity": "blockbuster/significant/moderate/niche", "market_dynamics_reasoning": "<2-3 sentences>" }}

Expected Output Format

This prompt expects a JSON response. See the user prompt template for the exact structure.

5

Step 5: Final Prediction

You are a senior biotech equity analyst making stock price predictions based on catalyst analysis.
Make your final stock price impact prediction based on all the analysis. STEP 1 - EXTRACTED FACTS: {step1_result} STEP 3 - CLINICAL ASSESSMENT: {step3_result} STEP 4 - MARKET DYNAMICS: {step4_result} CATALYST INFO: - Ticker: {ticker} - Company: {company} - Drug: {drug} IMPORTANT: Be realistic. Most FDA approvals after successful Phase 3 are neutral or slightly_positive because they're priced in. Return your final prediction: {{ "predicted_impact": "very_positive|positive|slightly_positive|neutral|slightly_negative|negative|very_negative", "score": <0-100>, "reasoning": "<2-3 sentence summary>", "key_factors": ["factor 1", "factor 2", "factor 3"] }}

Template Variables Reference

These variables are dynamically replaced with actual values when the strategy is executed:

{cat_type}Catalyst event type (e.g., FDA approval, trial results)
{company}Company name
{drug}Drug or therapy name
{indication}Medical indication/disease being treated
{phase}Clinical trial phase (1, 2, 3, etc.)
{pr_text}Full press release text
{step1_result}Output from Step 1 prompt
{step2_result}Output from Step 2 (web search)
{step3_result}Output from Step 3 prompt
{step4_result}Output from Step 4 prompt
{ticker}Company stock ticker symbol
{{ "clinical_significance": "breakthrough/meaningful/incremental/neutral/disappointing/failure", "vs_standard_of_care": "<comparison>", "unmet_need_addressed": "<assessment>", "approvability_assessment": "<likely/uncertain/unlikely>", "clinical_reasoning": "<2-3 sentences>" }Custom variable
{{ "predicted_impact": "very_positive|positive|slightly_positive|neutral|slightly_negative|negative|very_negative", "score": <0-100>, "reasoning": "<2-3 sentence summary>", "key_factors": ["factor 1", "factor 2", "factor 3"] }Custom variable
{{ "priced_in_assessment": "fully_priced_in/partially_priced_in/not_priced_in/negatively_positioned", "market_expectations": "exceeded/met/missed", "asset_importance": "lead_asset/key_pipeline/minor_program", "market_opportunity": "blockbuster/significant/moderate/niche", "market_dynamics_reasoning": "<2-3 sentences>" }Custom variable
{{ "primary_endpoint_met": true/false/null, "p_value": "<p-value if mentioned, else null>", "hazard_ratio": "<HR if mentioned, else null>", "confidence_interval": "<CI if mentioned, else null>", "effect_size": "<key efficacy metric>", "safety_signals": ["list of safety concerns mentioned"], "regulatory_status": "<FDA status or null>", "trial_phase": "<Phase 1/2/3 or null>", "patient_population": "<description of patients>", "key_quotes": ["1-2 important quotes"] }Custom variable