Run2 itinerary planning

Skill cxcscmu/SkillLearnBench/skills/b2-self-feedback-claude-sonnet-4-6/travel-planning/run2_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 run2_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

Build a multi-city road-trip travel itinerary from dataset files with complete budget tracking, pet-friendly constraints, cuisine coverage, and correct JSON output format.

SKILL.md

3.7 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it

Multi-City Road-Trip Itinerary Planning (Improved)

Planning Checklist

Step 1: Route and Day Assignment

Identify origin and N destination cities
Fetch driving distances for all pairs
Assign city days:
- Day 1: Origin → City 1 (driving day, ~full day)
- Days 2..k1: City 1 exploration
- Day k1+1: City 1 → City 2 (driving day)
- ...
- Last day: Last City → Origin (driving day, no accommodation needed)

Step 2: Accommodation Selection

Filters (all must pass):

city column matches target city (case-insensitive, exact city name)
house_rules does NOT contain "No pets"
maximum occupancy >= number of travelers
minimum nights <= number of nights planned

Sort preference: Higher review rate, lower price.

Step 3: Restaurant Selection

Use Cuisines column to match preferred cuisine types
Spread cuisine variety across days (American, Mediterranean, Chinese, Italian)
Avoid assigning restaurants from wrong city
Different restaurants per meal where possible

Step 4: Budget Calculation

total_cost = 0
for each city:
    total_cost += accommodation.price * nights_in_city

for each meal (not "-"):
    total_cost += restaurant.avg_cost * num_travelers

assert total_cost <= budget

Step 5: Attraction Selection

Use attractions.csv with exact city match
List multiple attractions on exploration days
Use "-" for driving-only days (unless short drive arrives early)

JSON Output Format Rules

{
  "plan": [
    {
      "day": 1,
      "current_city": "from Minneapolis to Cleveland",
      "transportation": "Self-driving: from Minneapolis to Cleveland, 11 hours 14 mins, 1,219 km",
      "breakfast": "Restaurant Name, City",
      "lunch": "-",
      "dinner": "Restaurant Name, City",
      "attraction": "-",
      "accommodation": "Accommodation NAME, City"
    }
  ],
  "data_sources": [...]
}

Field rules:

current_city: "from A to B" on transit days; city name on stay days
transportation: "Self-driving: from A to B, DURATION, DISTANCE" on transit days; "-" on stay days
attraction: semicolon-separated ending with ;; "-" on full driving days
accommodation: "NAME, City" using exact NAME from CSV; "-" on final return day
Meal fields: "Restaurant Name, City" or "-" if skipped
Lunch may be "-" on long driving days

Budget Example (2 people, 7 days, $5,100, 3 Ohio cities)

Item	Calculation	Cost
Cleveland accommodation (2 nights)	2 × $408	$816
Columbus accommodation (2 nights)	2 × $861	$1,722
Cincinnati accommodation (2 nights)	2 × $264	$528
Day 1 meals (2 people)	($14+$27) × 2	$82
Day 2 meals (2 people)	($36+$54+$80) × 2	$340
Day 3 meals (2 people)	($25+$65+$31) × 2	$242
Day 4 meals (2 people)	($24+$55+$70) × 2	$298
Day 5 meals (2 people)	($86+$67+$62) × 2	$430
Day 6 meals (2 people)	($26+$20+$93) × 2	$278
Day 7 meals (2 people)	$26 × 2	$52
TOTAL		$4,788

Budget remaining: $312 (within $5,100 budget ✓)

Cuisine Coverage Table

Track coverage to ensure each preferred cuisine appears multiple times:

Cuisine	Meals
American	Day2-lunch, Day4-breakfast, Day5-breakfast, Day6-breakfast, Day7-breakfast
Mediterranean	Day1-breakfast, Day2-breakfast, Day3-lunch, Day5-dinner
Chinese	Day3-breakfast, Day4-lunch, Day5-lunch, Day6-dinner
Italian	Day1-dinner, Day2-dinner, Day3-dinner, Day4-dinner, Day6-lunch