Compare HotelsCAUTION

AI-powered travel hacking and search with cash, points, miles, and award flights. Drop-in skills and MCP servers for Claude, Codex, and OpenCode.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: compare-hotels
description: Unified hotel comparison across Chase Edit, Amex FHR/THC, metasearch, and Airbnb. Highlights premium program benefits, stacking opportunities, and benefit-adjusted pricing.
category: orchestration
summary: Unified hotel comparison across portals, metasearch, and Airbnb. FHR/Edit stacking detection.
api_key: Uses individual skill keys
---

**Companion reference skills.** Load these for deeper context:
- `hotel-chains` — branded property to loyalty program mapping
- `premium-hotels` — FHR / THC / Chase Edit credits and stacking strategy
- `points-valuations` — CPP floor/ceiling for hotel programs (Hyatt 1.4-1.7cpp, Hilton 0.4cpp floor, etc.)
- `fallback-and-resilience` — recovery when a hotel source fails


# Compare Hotels

Search every available source for accommodation and present one unified comparison. Combines portal premium programs (Chase Edit, Amex FHR/THC), cash pricing, and Airbnb alternatives into a single view.

**This is an orchestration skill.** It tells the agent which tools to run and how to combine results. No standalone script.

## When to Use

- "Find me a hotel in Paris for Aug 11-15"
- "Should I book through Chase or Amex for this hotel?"
- "Compare hotels vs Airbnb in Oslo"
- Any accommodation search where the user wants the full picture

## Sources

Run these in parallel where possible. **Never fail silently.**

### Portal Premium Programs (Docker required)

| Source | Skill | Speed | What It Finds |
|--------|-------|-------|---------------|
| Chase Travel | `chase-travel` | ~45s | Edit hotels ($100 credit, breakfast, upgrade), Points Boost, standard hotels |
| Amex Travel | `amex-travel` | ~45s | FHR ($100+ credit, breakfast, upgrade, late checkout), THC ($100 credit on 2+ nights) |

### Cash Prices and Metasearch

| Source | Skill/Tool | Speed | What It Finds |
|--------|------------|-------|---------------|
| SerpAPI | `serpapi` | ~3s | Google Hotels cash prices and direct booking links. |
| Trivago | web search | ~5s | Metasearch across many OTAs. Use tavily/exa to search `trivago.com`. Finds the absolute cheapest OTA rate. |
| LiteAPI | web search | ~5s | Hotel API aggregator with negotiated rates. Check `liteapi.travel`. Can find rates below public OTA pricing. |
| RapidAPI | `rapidapi` | ~3s | Booking.com inventory and pricing. Optional, use when user asks for Booking.com specifically. |

### Quality and Ratings

| Source | Skill/Tool | Speed | What It Finds |
|--------|------------|-------|---------------|
| TripAdvisor | `tripadvisor` | ~2s | Ratings, rankings, subratings (location/service/cleanliness), reviews, photos, amenities, awards. 5K calls/month. |

**SerpAPI, Trivago, and LiteAPI are all default.** Always search all three. Only add RapidAPI if the user asks for Booking.com specifically. Use TripAdvisor for quality data (ratings, rankings, reviews) to complement pricing from other sources.

### Vacation Rentals

| Source | Skill/Tool | Speed | What It Finds |
|--------|------------|-------|---------------|
| Airbnb | `airbnb_search` MCP tool | ~5s | Entire homes, private rooms, with total pricing |
| VRBO | `vrbo` skill (Patchright) | ~20s | Whole homes, condos, cabins. Professionally-managed units often not on Airbnb. Headed/Docker — behind Akamai. |

### Premium Property Databases (local, instant)

| Source | Data File | What It Finds |
|--------|-----------|---------------|
| FHR properties | `data/fhr-properties.json` | Check if a hotel is FHR before booking |
| THC properties | `data/thc-properties.json` | Check if a hotel is THC |
| Chase Edit properties | `data/chase-edit-properties.json` | Check if a hotel is in Chase's Edit program |

## Workflow

### Step 1: Search All Sources

```
PARALLEL GROUP 1 (fast — HTTP/API/MCP calls, no browser, ~3-5s):
  - SerpAPI: Google Hotels cash prices
  - Trivago: metasearch across OTAs (web search)
  - LiteAPI: negotiated hotel rates (web search)
  - Airbnb: vacation rentals in the area (npx MCP — lightweight scrape, not a browser)
  - Local data: check premium-hotels databases for the city

PARALLEL GROUP 2 (slow — headed browser via Patchright/Xvfb in Docker, ~20-45s):
  - VRBO: whole homes / condos / cabins — vrbo skill (~20s, headed/Docker, behind Akamai; run for whole-home comparisons alongside Airbnb)
  - Chase Travel: --hotel --dest "City" --checkin YYYY-MM-DD --checkout YYYY-MM-DD --json
  - Amex Travel: --hotel --dest "City" --checkin YYYY-MM-DD --checkout YYYY-MM-DD --json
```

For each source, capture:
- Source name
- Status: "ok" | "error: {reason}" | "skipped: {reason}"
- Results count
- Data

### Step 2: Identify Stacking Opportunities

Cross-reference results with `data/fhr-properties.json`, `data/thc-properties.json`, and `data/chase-edit-properties.json`:

```
For each hotel in results:
  - Is it in FHR? → Tag [FHR]
  - Is it in THC? → Tag [THC]
  - Is it in Chase Edit? → Tag [EDIT]
  - Is it in multiple? → Tag [FHR+EDIT] and note stacking potential
```

Stacking means: book through one portal, pay with the other card separately. For example, book FHR through Amex (get breakfast + credit + upgrade), then use CSR $300 travel credit on something else at the property.

### Step 3: Calculate Points Value

**Chase:**
- Chase portal pricing on CSR/CSP is dynamic Points Boost. Each booking quotes a specific points price; effective cpp typically falls in the 1.5-2.0 cpp range on CSR but is not a guaranteed floor on every booking. Pull the actual portal quote.
- Points Boost offers are dynamic and vary by route/date/inventory.
- The portal quote IS the actual UR cost. Don't divide by 1.5 — the listed points already reflect the dynamic Boost rate at quote time.

**Amex:**
- Hotel portal pricing is variable (NOT always 1 cpp like flights).
- The points price shown is the actual MR cost. Calculate CPP from cash/points ratio.
- FHR benefits ($100+ credit, breakfast) offset the rate significantly.

**Calculate benefit-adjusted price:**
```
FHR adjusted = total_pric

CAUTIONgrade B · trust 89/100 Install with care. The audit found things worth knowing before you trust its output.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (5)

.agents/skills

.claude/skills

.opencode/skills

skills

AGENTS.md

Gates applied: no_behavioural_pass.