Itinerary planner
[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.
npx -y skills add cxcscmu/SkillLearnBench --skill itinerary-plannerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
SKILL.md
1.2 KB, 290 tokens by cl100k_base, as published. Nobody here has run it
name: itinerary-planner description: How to construct a 7-day travel itinerary with constraints. Use this skill when organizing days, managing budget, and ensuring all user requirements (pet-friendly, cuisine, no flights) are met.
Planning Workflow
1. Route Optimization
- Start from the origin (Minneapolis).
- Sequence cities to minimize backtracking and excessive driving on a single day (except for the long hauls from/to Minneapolis).
- Use
distance.csvto validate the route.
2. Constraint Checklist
- 7 Days: Ensure exactly 7 day objects in the final JSON.
- Three Cities: Cover the requested number of target cities.
- Pet-Friendly: Every accommodation MUST have a rule allowing pets.
- Cuisines: Incorporate preferred cuisines (American, Mediterranean, Chinese, Italian) across the meals.
- Budget: Keep total cost (Accommodations + Meals + Estimated Fuel) under the limit ($5,100).
- Fuel Estimate: $0.15 per km (roughly).
- Meal Estimate: Use
Average Costfrom restaurant data.
3. Output Format
- Strictly follow the JSON structure provided in the task description.
- Ensure
current_cityaccurately reflects the location or the movement (e.g., "from A to B"). attractionfield must end with;.