agentsclimarketplace

Run3 itinerary validation

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

[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-validation

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 validate a completed itinerary JSON against all task requirements before finalizing output. Run this checklist after formatting and before writing the file.

SKILL.md

2.3 KB, 536 tokens by cl100k_base, as published. Nobody here has run it

Itinerary Validation Checklist

Run every check below. If any check fails, revise the itinerary before writing the output file.

Structure Checks

  • plan array contains exactly 7 day objects
  • day values are integers 1 through 7 with no gaps or duplicates
  • data_sources array is present and lists all files actually queried

Travel Checks

  • ✅ No flights used anywhere in the plan (all transportation is self-driving or "-")
  • ✅ Driving days use "from A to B" format in current_city
  • ✅ Driving days use "Self-driving: from A to B" in transportation
  • ✅ Stay days use "-" in transportation
  • ✅ Route is logically ordered (origin → city 1 → city 2 → city 3 → return or similar)

City Checks

  • ✅ Exactly three Ohio cities are covered
  • ✅ All cities confirmed as valid in background/citySet_with_states.txt
  • ✅ Origin city is Minneapolis

Date and Duration Checks

  • ✅ Plan spans March 17–23, 2022 (7 days)

Accommodation Checks

  • ✅ All accommodations are confirmed pet-friendly from the dataset
  • ✅ No accommodation is "-" except on the final return-home day (if applicable)

Meal Checks

  • ✅ At least one meal per day is assigned (not all three are "-")
  • ✅ Cuisine preferences (American, Mediterranean, Chinese, Italian) appear across the plan
  • ✅ All restaurants are sourced from restaurants/clean_restaurant_2022.csv for the correct city

Attraction Checks — CRITICAL

  • Day 1 attraction is not empty or "-" ← Day 1 is the most commonly missed; verify explicitly
  • ✅ Every day (Days 1–7) has a non-empty, non-"-" attraction string
  • ✅ Every attraction string ends with ;
  • ✅ Attractions are sourced from attractions/attractions.csv for the correct city
  • ✅ No day uses "-" for attraction, even if it is a driving/travel day

Budget Check

  • ✅ Total estimated cost ≤ $5,100 for two travelers
  • ✅ Budget breakdown covers: driving costs, accommodation, meals, and attraction entry fees

Final Output Check

  • ✅ JSON is valid (no syntax errors)
  • ✅ File written to /app/output/itinerary.json

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.