Parametric Casualty ILS Trigger Evaluator with Claude Agents SDK and Extended Thinking
Casualty Insurance-Linked Securities (ILS) require transparent, auditable parametric triggers based on multi-party legal or market loss indexes. You will build a parametric trigger monitor using the Claude Agents SDK with Extended Thinking enabled to evaluate complex loss index feeds against bond payout conditions. The agent must yield zero false triggers and maintain auditable reasoning traces.
What you are building
The core problem, expected build, and operating context for this challenge.
Construct an auditable parametric ILS trigger agent using Claude Agents SDK and Extended Thinking capabilities to process capital market loss indexes.
How work is evaluated
Evaluates parametric ILS trigger evaluation accuracy across 50 casualty market index scenarios.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
payout_math_exact
Payout amount must exactly match linear interpolation formula.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
zero_false_trigger_rate
1.0 if zero un-triggered claims activated payout. • target: 1 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Initialize Claude Agents SDK agent configured with extended thinking budget parameters.
Design custom triggers for Casualty ILS contracts based on industry loss indicators.
Implement multi-stage deterministic audit tools linked into the Claude Agent reasoning loop.
Verify indemnity vs parametric payout calculations with exact financial accounting.
Reference links and supporting material
Dataset of 100 Casualty ILS bond prospectuses detailing attachment points, exhaustion thresholds, and historical industry loss index movements.
Challenge setup
Evaluates parametric ILS trigger evaluation accuracy across 50 casualty market index scenarios.
Challenge input
JSON containing bond_id, attachment_point_usd, exhaustion_point_usd, reporting_index_value_usd
Claude Agents SDK
Offers reasoning and extended thinking capabilities for financial contract analysis.
Evaluated output
JSON containing trigger_activated (BOOLEAN), principal_loss_percent, payout_amount_usd, thinking_summary
- Payout amount must exactly match linear interpolation formula.
- 1.0 if zero un-triggered claims activated payout. • target: 1 • range: 0-1
- Benchmark: ILSTriggerEval-2026
- Zero False Trigger Rate target: 100%
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Featherless AI · Required
- Paymon · Optional
- DeepSeek V3 · Optional
Evaluation contract
- payout_math_exact · Weight 1
- zero_false_trigger_rate · Weight 1
Metrics
- zero_false_trigger_rate · Target 1
Benchmarks
- ILSTriggerEval-2026
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing bond_id, attachment_point_usd, exhaustion_point_usd, reporting_index_value_usd
Output: JSON containing trigger_activated (BOOLEAN), principal_loss_percent, payout_amount_usd, thinking_summary
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.