agentleFS
mtnrabimtnrabi/travel-agent-skills/skills/destination-compare/SKILL.md

destination-compare

Skill4 starsChanged 24 days agoMITView on GitHub

Pick WHERE to go - price many candidate destinations from one origin over the same date window with real-time Google Flights fares and rank the cities cheapest first. Use this when the destination is still open and the dates are roughly set, the inverse of cheapest-dates. Fans a LIST of to_airport IATA codes out in a single MCP call (search_oneway_flights and search_roundtrip_flights take a destination list plus departure_date_from, departure_date_to and nights), or one REST POST per city on google-flights-live-api. Use when the user asks where can I go, where should I go, anywhere, surprise me, somewhere warm, sunny, beach, island or ski, cheap flights anywhere, flight deals from JFK, LHR, LAX, TLV or my home airport, cheapest place, city or country to fly to, explore the map, destination ideas, weekend trip ideas, best long-weekend or city-break destination on a budget, where can I go for under $300 or under EUR 500 in October, compare or shortlist fares to several cities (Rome vs Athens vs Lisbon), which of these countries or islands is cheapest to reach, or is it cheaper to fly to X or Y. Returns a price-ranked table - cheapest fare per city with price_as_number or total_price_as_number, airline, stops, duration, departure_date and a bookable buy_link - plus a coverage line naming every destination and date actually searched, because the shortlist is one the agent chose, not a scan of every airport on earth. Destination already named and only the date in question - use cheapest-dates. Destination named and a hotel to add into one budget - use trip-planner. One fixed route re-checked over time - use fare-watch. Fares are fetched live, never cached. Not affiliated with Google.

  • Reads credentials
  • Sends data out

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read search_coverage and report it.

    line 29
  2. Very specific:

    Always close with a coverage line built from search_coverage: which destinations and dates were actually searched, whether the run was …

    line 99
  3. Distinctive:

    If this file names another skill from this set (travel-data-api, cheapest-dates, destination-compare, trip-planner, fare-watch, …

    line 103

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Destination Compare

Overview

This skill answers open-ended "where can I go" questions: one origin, one date window, many candidate destinations, ranked by real fare. Auth is one header, x-rapidapi-key, read from the RAPIDAPI_KEY environment variable and sent only to the FlightPowers hosts named in this file (full reference: the travel-data-api skill, or https://flightpowers.com/docs). Prefer the hosted MCP tools when a client has them configured (ad-free: https://flights.flightpowers.com/mcp and https://hotels.flightpowers.com/mcp) — search_oneway_flights and search_roundtrip_flights accept a LIST of destinations and a date RANGE, so a 10-city comparison is ONE tool call instead of N REST calls. The fields you rank on are price_as_number (one-way) and total_price_as_number (round-trip); the fields you report coverage from are search_coverage.destinations_searched, search_coverage.departure_dates_searched, and search_coverage.truncated.

Endpoints and Tools

JobCall
Many destinations, one-way, date rangeMCP search_oneway_flights — to_airport as a list, departure_date_from / departure_date_to
Many destinations, round-trip, flexible trip lengthMCP search_roundtrip_flights — to_airport as a list, nights as a number or list
No MCP client configured — one destination per callPOST /api/google_flights/oneway/v1 on google-flights-live-api.p.rapidapi.com
Round-trip over RESTPOST /api/google_flights/roundtrip/v1 on the same host
Add nightly hotel cost to a shortlisted cityPOST /search on booking-live-api.p.rapidapi.com — see the hotel-search skill

Workflow

  1. Pin down origin, window, budget, and trip shape before calling anything. If the user gave a month ("October"), pick a concrete sub-range — a full month across 10 destinations is 310 combinations and will be truncated.
  2. Build the candidate destination list yourself, as IATA codes. The API does not suggest destinations; it prices the ones you name. Say which cities you picked and why before spending the user's money.
  3. Count the combinations (destinations x departure dates) and state the cost out loud. Cap: with the caller's own key the MCP server searches up to 30 combinations per call by default and raises that by itself for a bigger request, up to 300 (max_searches sets it explicitly); on the free sign-in allowance (10 searches a day, no key) a call larger than what is left of the day is refused before it runs (or sampled down to it when max_searches is set). Past the cap the server samples evenly and sets search_coverage.truncated, so split into batches by theme ("Greek islands", "Balkans") rather than silently dropping cities.
  4. Make ONE MCP call with the whole destination list. Never loop one call per destination or per date when an MCP client is available.
  5. Rank client-side by price_as_number (or total_price_as_number), keeping the single cheapest row per destination plus its date, airline, stops and buy_link.
  6. Read search_coverage and report it. A city that returned no rows is "no fare found in this window", never "there are no flights".

Single Comparison Run (REST, one destination)

Runnable as written once RAPIDAPI_KEY is set in the environment.

: "${RAPIDAPI_KEY:?set RAPIDAPI_KEY to your FlightPowers key}"

curl -s -X POST \
  "https://google-flights-live-api.p.rapidapi.com/api/google_flights/oneway/v1" \
  -H "x-rapidapi-key: $RAPIDAPI_KEY" \
  -H "x-rapidapi-host: google-flights-live-api.p.rapidapi.com" \
  -H "content-type: application/json" \
  -d '{"from_airport":"TLV","to_airport":"ATH","departure_date":"2027-06-21","currency":"USD","max_price":300,"limit":10}' \
  | jq '[.[] | {to_airport, departure_date, price_as_number, airline, stops, buy_link}] | sort_by(.price_as_number)'

Comparing Many Destinations

With an MCP client, one call covers the whole shortlist:

{
  "tool": "search_oneway_flights",
  "arguments": {
    "from_airport": "TLV",
    "to_airport": ["ATH", "LCA", "SOF", "BUD", "TIA", "SKG"],
    "departure_date_from": "2027-06-21",
    "departure_date_to": "2027-06-23",
    "currency": "usd",
    "max_price": 300,
    "limit": 40
  }
}

That is 6 destinations x 3 dates = 18 combinations: 18 billed searches on the caller's own key, inside the default 30-per-call cap. On the free sign-in allowance it would be refused (more than the day's 10), so offer 3 cities x 3 dates instead if the day's allowance is untouched (check trial.remaining_today in the last response). For round-trip, swap the tool for search_roundtrip_flights and pass "nights": [3, 4] instead of a return date.

Without an MCP client, REST needs one request per destination per date. Loop explicitly and keep the per-destination minimum:

: "${RAPIDAPI_KEY:?set RAPIDAPI_KEY to your FlightPowers key}"

for DEST in ATH LCA SOF BUD TIA SKG; do
  curl -s -X POST \
    "https://google-flights-live-api.p.rapidapi.com/api/google_flights/oneway/v1" \
    -H "x-rapidapi-key: $RAPIDAPI_KEY" \
    -H "x-rapidapi-host: google-flights-live-api.p.rapidapi.com" \
    -H "content-type: application/json" \
    -d "{\"from_airport\":\"TLV\",\"to_airport\":\"$DEST\",\"departure_date\":\"2027-06-21\",\"currency\":\"USD\",\"max_price\":300,\"limit\":10}" \
    | jq -c --arg dest "$DEST" '{dest: $dest, offers: length, cheapest: (map(.price_as_number) | min)}'
done

cheapest comes back null for a destination with no offers — that is the honest "searched, found nothing" case, and it is exactly what you must show instead of omitting the row.

Common Pitfalls

  • An empty result set is an answer only when the search completed. REST returns a bare [] with HTTP 200 and the MCP tools return "results": [] with result_count: 0. Check X-Search-Status on REST: ok/empty means "no flights on this route and date under these filters" — report it per destination and move on. degraded means the search did not complete, so that destination is unknown, not unreachable. Never let a failed search become "you cannot get there from here".
  • limit caps the merged, price-sorted list, so one cheap city or date can crowd all the others out. A date or city can be in search_coverage.departure_dates_searched / destinations_searched and still have no row in results, because cheaper rows from elsewhere filled the list. On the MCP tools by_destination carries one entry per requested destination with a reason that separates no flights, not searched, a failed search, and fares that did not fit in limit; read it before writing a city off. Never read "no rows for X" as "no flights to X". Raise limit for wide comparisons, or run a confirming single-destination call before writing a city or a date off.
  • Ranking destinations against each other is client-side work. The API sorts within one search, never across them. Do not pass sort_type to the MCP tools; it is not one of their parameters (they take sort_by: best, price or duration).
  • If a destination comes back degraded, a plain retry is the right move, announced, because it is another billed request. Leave use_fallback unset: unset already lets a search whose own retries failed escalate once to a second data source, while true forces that source on every attempt, which is much slower and can time out.
  • Every combination is a billed request. Destinations x departure dates is the multiplier, and it is the user's money. Quote the number of searches before you run them, not after. On the MCP servers, api_usage in the response tells you what the call actually cost.
  • Never call a fare cheap without the field backing it. price_range_in_relation_to_other_periods (low | typical | high) and price_insights_low / price_insights_high are what let you say a fare is a good deal. Read the verdict off the row you are quoting; if it is missing or empty for that row, quote the fare and nothing more. Do not infer "low" from the fact that a city sorted to the top of your comparison — that only means it beat the other cities on your shortlist.
  • If you extend the comparison with hotel cost, available: false with null prices means sold out for those dates. That is a valid answer, not a failure; show the city as "no availability for these dates" rather than dropping it from the table.

Output Standards

  • Lead with the ranked table — destination, cheapest fare, date, airline, stops, duration, buy_link — cheapest first, then say which cities came in over budget or returned nothing.
  • Always close with a coverage line built from search_coverage: which destinations and dates were actually searched, whether the run was truncated, and the fact that the shortlist was yours, not an exhaustive scan of every airport on earth.
  • Stamp every table with the fetch time and treat the fares as valid for minutes. Never reuse or cache an earlier comparison; re-run the search when the user comes back.
  • Non-affiliation: this is an independent API returning publicly available flight and hotel pricing. It is not affiliated with, endorsed by, or sponsored by Google or Booking.com.

If this file names another skill from this set (travel-data-api, cheapest-dates, destination-compare, trip-planner, fare-watch, hotel-search, rate-parity-monitor, travel-brief) that is not installed, do that part of the job with the endpoints in this file.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.