LifeSG Multi-Agency Citizen Service Assistant using OpenAI Agents SDK
Navigating multi-agency government benefits across major life milestones in Singapore can be challenging for families seeking personalized assistance. Build a multi-turn citizen assistant using the OpenAI Agents SDK to orchestrate LifeSG service workflows, assess eligibility for family grants, and execute multi-step government transaction applications securely.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an autonomous citizen service orchestrator using OpenAI Agents SDK with function calling and handoffs to streamline LifeSG multi-agency transactions.
How work is evaluated
Evaluates OpenAI Agents SDK handoff efficiency, tool call accuracy, and grant calculation correctness across multi-turn user 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.
grant_total_check
Verifies that total calculated grant matches official policy amounts for 2nd child
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
handoff_precision
Accuracy of agent routing based on user intent • 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
Construct multi-agent handoff workflows using OpenAI Agents SDK
Integrate custom function calling tools for Singpass profile and grant lookup
Evaluate eligibility across multiple Singapore government schemes (Baby Bonus, CDA, Support Grants)
Ensure privacy-preserving data handling in citizen service interactions
Reference links and supporting material
Synthetic citizen profile records and Singapore public grant scheme eligibility schedules.
Challenge setup
Evaluates OpenAI Agents SDK handoff efficiency, tool call accuracy, and grant calculation correctness across multi-turn user scenarios.
Challenge input
JSON containing household income, child birth rank, and citizenship status
OpenAI Agents SDK
OpenAI's official framework for multi-agent handoffs, stateful threads, and function calling
Evaluated output
JSON containing eligible_schemes, total_grant_value_sgd, and agent_handoffs_executed
- Verifies that total calculated grant matches official policy amounts for 2nd child
- Accuracy of agent routing based on user intent • target: 1 • range: 0-1
- Benchmark: Autonomous Citizen Workflow Benchmark
- Handoff Precision target: 1
- 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
- OpenAI · Required
- Zed · Optional
- CleverBee · Optional
Evaluation contract
- grant_total_check · Weight 1
- handoff_precision · Weight 1
Metrics
- handoff_precision · Target 1
Benchmarks
- Autonomous Citizen Workflow Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing household income, child birth rank, and citizenship status
Output: JSON containing eligible_schemes, total_grant_value_sgd, and agent_handoffs_executed
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Orchestration Framework
Recipe freeze marker
No recipe freeze time is recorded.