Create precise function definitions that trigger in the RIGHT scenarios and stay silent in the WRONG ones. You're competing against other functions for attention.
š HOW TO PLAY
1. Review the test scenarios - some should trigger your function (ā SHOULD), others shouldn't (ā SHOULD NOT)
2. Write your function definition with a clear name and description
3. Add parameters with proper types and descriptions (optional but helps)
4. Use enums to constrain values when appropriate
5. Test your function against all scenarios
š§ FUNCTION SCHEMA FORMAT
Your function follows the JSON Schema / OpenAPI format:
name: Function identifier (e.g., "get_weather")
description: What your function does - BE SPECIFIC
parameters: Object with type definitions
properties: Define each parameter with type, description
required: Array of required parameter names
enum: Use to limit values (e.g., ["celsius", "fahrenheit"])
š” PRO TIPS
Names matter: Specific names beat generic ones. "fetch_user_profile" > "get_data"
Descriptions are key: AI reads your description to decide. Be clear about WHAT and WHEN.
Use enums wisely: They constrain inputs and make intent clearer.
Think like AI: What keywords would make YOUR function the obvious choice?
š SCORING
+100 points for each correct trigger (when you SHOULD)
-50 points for each incorrect trigger (when you SHOULDN'T)
+50 bonus points for each FALSE scenario where you correctly didn't trigger
Compete against 0-5 competitor functions trying to steal your scenarios!
š LEARNING PATH
Levels progress from basic scenarios to complex edge cases:
ā¢ Novice: Simple, clear scenarios
ā¢ Intermediate: Ambiguous queries, need precision