Run3 itinerary planning

Skill cxcscmu/SkillLearnBench/skills/b3-teacher-feedback-claude-sonnet-4-6/travel-planning/run3_itinerary-planning

[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.

Install

npx -y skills add cxcscmu/SkillLearnBench --skill run3_itinerary-planning

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Use this skill to build a multi-day travel itinerary from structured requirements (origin, cities, dates, budget, constraints). It queries available database files and produces a day-by-day plan with transportation, meals, attractions, and accommodations.

SKILL.md

2.8 KB, 615 tokens by cl100k_base, as published. Nobody here has run it

Itinerary Planning Skill

Step 1: Parse Requirements

Extract the following from the user's request:

Origin city (e.g., Minneapolis)
Destination cities (e.g., three Ohio cities)
Travel dates (start date → end date, total days)
Number of travelers
Budget (total, in USD)
Constraints: no flights, pet-friendly, cuisine preferences, etc.

Step 2: Select Cities and Routing

Query background/citySet_with_states.txt to confirm valid Ohio cities (e.g., Cleveland, Columbus, Cincinnati).
Query googleDistanceMatrix/distance.csv to find driving distances and durations between Minneapolis and each Ohio city, and between Ohio cities.
Choose a logical driving route that minimizes total driving time and fits within 7 days.
Assign driving days (typically Day 1 and one or two mid-trip transition days).

Step 3: Assign Daily Structure

For each of the 7 days, assign:

current_city: use "from A to B" format on driving days
transportation: "Self-driving: from A to B" on driving days; "-" on stay days
breakfast, lunch, dinner: select from restaurants/clean_restaurant_2022.csv filtered by city and preferred cuisines
attraction: Always assign at least one attraction per day, including Day 1 and all other travel/driving days. On driving days, include a stop en route or an attraction at the destination city upon arrival. Never leave attraction as "-" or empty.
accommodation: select from accommodations/clean_accommodations_2022.csv filtered by city and pet-friendly flag; use "-" only on the final departure day if the traveler returns home

Day 1 Special Rule

Day 1 is a driving day from the origin to the first destination city. It must still contain at least one attraction — either a notable stop along the driving route, or an attraction visited upon arriving at the destination city. Do not leave Day 1 attraction empty.

Step 4: Budget Tracking

Estimate costs across all 7 days:

Transportation: use distance × per-mile cost estimate (~$0.18/mile for self-driving)
Accommodation: nightly rate × nights from the accommodations dataset
Meals: average meal cost × number of meals × 2 travelers
Attractions: entry fees where applicable
Confirm total ≤ stated budget (e.g., $5,100)

Step 5: Pet-Friendly Filter

All accommodations must be flagged as pet-friendly in the dataset.
Do not select any hotel/motel without confirmed pet-friendly status.

Step 6: Output

Produce a valid JSON object matching the required output schema (see task Output Format).