Skip to content
Back to skills

Travel Planner

ASecurity

A full pipeline where an agent team collaborates to generate everything from destination analysis to local information for travel planning. Used for 'plan a trip', 'Japan travel itinerary', 'Europe backpacking', 'weekend getaway', 'international travel prep', 'travel budget', 'accommodation recommendations', 'travel itinerary', 'local restaurants', 'travel route recommendations', etc. Supports analysis or improvement of existing itineraries. Real-time flight/accommodation booking, visa applic...

  • 3 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added May 27, 2026
businessgo

Security analysis

A100/100

Scanned May 27, 2026

npx -y skills add chodolmu/taskforge-pro --skill travel-planner --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Travel Planner?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Travel Planner
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/chodolmu-travel-planner/badge)](https://www.skillsdirectory.com/skills/chodolmu-travel-planner)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: travel-planner
description: "A full pipeline where an agent team collaborates to generate everything from destination analysis to local information for travel planning. Used for 'plan a trip', 'Japan travel itinerary', 'Europe backpacking', 'weekend getaway', 'international travel prep', 'travel budget', 'accommodation recommendations', 'travel itinerary', 'local restaurants', 'travel route recommendations', etc. Supports analysis or improvement of existing itineraries. Real-time flight/accommodation booking, visa application processing, and travel agency package comparisons are outside this skill's scope."
---

# Travel Planner — Travel Planning Full Pipeline

Generates destination analysis, itinerary, accommodation, budget, and local information through agent team collaboration in one go.

## Execution Mode

**Agent Team** — 4 agents communicate directly via SendMessage and cross-verify.

## Agent Composition

| Agent | File | Role | Type |
|-------|------|------|------|
| destination-analyst | `.claude/agents/destination-analyst.md` | Destination research, attractions, safety info | general-purpose |
| itinerary-designer | `.claude/agents/itinerary-designer.md` | Daily itinerary, route optimization, accommodation | general-purpose |
| budget-manager | `.claude/agents/budget-manager.md` | Budget calculation, cost comparison, savings tips | general-purpose |
| local-guide | `.claude/agents/local-guide.md` | Transport, restaurants, culture, emergency info | general-purpose |

## Workflow

### Phase 1: Preparation (Orchestrator performs directly)

1. Extract from user input:
    - **Destination**: Country/City (recommend based on criteria if undecided)
    - **Duration**: How many days, departure/arrival dates
    - **Party**: Solo/Couple/Family/Friends + number of people
    - **Budget** (optional): Total budget or budget level (Budget/Standard/Luxury)
    - **Travel Style** (optional): Relaxation/Adventure/Cultural/Culinary
    - **Special Requests** (optional): Must-visit places, exclusions
2. Create `_workspace/` directory at project root
3. Organize inputs and save to `_workspace/00_input.md`
4. **Determine execution mode** based on request scope

### Phase 2: Team Assembly and Execution

| Order | Task | Owner | Dependencies | Deliverable |
|-------|------|-------|-------------|-------------|
| 1 | Destination Analysis | analyst | None | `_workspace/01_destination_analysis.md` |
| 2 | Itinerary + Accommodation | designer | Task 1 | `_workspace/02_itinerary.md`, `_workspace/03_accommodation.md` |
| 3a | Budget Plan | budget | Tasks 1, 2 | `_workspace/04_budget.md` |
| 3b | Local Information | guide | Tasks 1, 2 | `_workspace/05_local_guide.md` |

Tasks 3a (Budget) and 3b (Local Info) run **in parallel**.

**Inter-team communication flow:**
- analyst complete → transmit attractions/time/hours to designer, costs/rates to budget, culture/safety to guide
- designer complete → transmit daily costs/accommodation budget to budget, daily areas/segments to guide
- budget complete → transmit payment info to guide
- guide complete → if inconsistencies found with itinerary, request adjustment from designer

### Phase 3: Integration and Final Deliverables

1. Verify all files in `_workspace/`
2. Confirm consistency across itinerary-budget-local information
3. Report final summary to user:
    - Destination Analysis — `01_destination_analysis.md`
    - Itinerary — `02_itinerary.md`
    - Accommodation Guide — `03_accommodation.md`
    - Budget Plan — `04_budget.md`
    - Local Guide — `05_local_guide.md`

## Modes by Request Scale

| User Request Pattern | Execution Mode | Agents Deployed |
|---------------------|---------------|-----------------|
| "Plan an entire trip" | **Full Pipeline** | All 4 |
| "Where should I go in Japan?" | **Destination Recommendation** | analyst solo |
| "Paris 3-night itinerary only" | **Itinerary Mode** | analyst + designer |
| "Just calculate travel budget" | **Budget Mode** | budget solo |
| "Info I need while there" | **Guide Mode** | guide solo |
| "Improve this itinerary" (existing file) | **Analysis Mode** | designer + budget |

**Using existing files**: If the user provides an existing itinerary, copy to `_workspace/02_itinerary.md` and skip the designer.

## Data Transfer Protocol

| Strategy | Method | Purpose |
|----------|--------|---------|
| File-based | `_workspace/` directory | Store and share major deliverables |
| Message-based | SendMessage | Real-time key info transfer, revision requests |
| Task-based | TaskCreate/TaskUpdate | Progress tracking, dependency management |

File naming convention: `{order}_{agent}_{deliverable}.{ext}`

## Error Handling

| Error Type | Strategy |
|-----------|----------|
| Web search failure | Work from general knowledge, note "Latest information verification needed" |
| Destination undecided | Recommend 3 destinations based on user criteria |
| Budget exceeded | Present budget alternatives + adjust priorities |
| Agent failure | Retry once → proceed without that deliverable if failed, note omission in report |
| Itinerary-info inconsistency | Request adjustment from designer (up to 2 times) |

## Test Scenarios

### Normal Flow
**Prompt**: "Plan a 4-night 5-day trip to Tokyo for a couple. Budget $1,500, focusing on food tourism."
**Expected Results**:
- Destination analysis: Tokyo attractions, culinary spots, entry requirements, optimal season
- Itinerary: 5-day schedule, food-focused routes, organized by areas like Tsukiji/Shibuya/Shinjuku
- Accommodation: 3 centrally-located recommendations + booking tips
- Budget: Allocation within $1,500 (flights/accommodation/meals/transport), Budget/Standard/Luxury comparison
- Local guide: Transit (Suica), restaurant list, basic Japanese phrases, emergency contacts

### Existing File Flow
**Prompt**: "Check if this itinerary fits the budget and add local info" + attached itinerary file
**Expected Results**:
- Copy existing itinerary to `_workspace/02_itinerary.md`
- Deploy budget + guide: generate budget calculation + local information

### Error Flow
**Prompt**: "Where should I go in Southeast Asia, budget under $800, solo for 3 nights 4 days"
**Expected Results**:
- Destination undecided → analyst compares 3 destinations based on criteria (Vietnam/Thailand/Philippines, etc.)
- Provide comparison table of estimated costs, pros/cons for each
- Proceed with full pipeline after user selection (or auto-proceed with top recommendation)

## Agent Extension Skills

| Agent | Extension Skill | Purpose |
|-------|----------------|---------|
| itinerary-designer | `route-optimizer` | Route optimization, transport comparison, time block allocation |
| budget-manager | `budget-calculator` | City-specific cost tables, budget allocation formulas, savings strategies |

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.