srt
This skill manages Korean SRT search, reservation, listing, and cancellation through local CLI tools and the SRT library. It requires SRT_PHONE/SRT_PASSWORD, executes local commands (e.g., uv run --with SRTrain python3 scripts/srt_cli.py ..., exec reserve --retry), and invokes srt.search_train/srt.reserve which contact external services.
SRT Korean Train Service Skill
OpenClaw skill for managing Korean SRT (Super Rapid Train) reservations with search, booking, view, and cancellation capabilities.
Features
- 🔍 Search trains between stations with real-time seat availability
- 🎫 Make reservations with automatic rate limiting to protect your account
- 📋 View bookings to see all active reservations
- 🗑️ Cancel bookings with confirmation prompts
- 🤖 AI-friendly JSON output for programmatic access
- 🛡️ Rate limiting to prevent account blocking (3s between reservations, 5s between searches)
- ⚠️ Retry protection with maximum 10 attempts per session
Prerequisites
- Python 3.10+ installed
- SRT account with phone number and password
- OpenClaw installed and configured
Configuration
Set your SRT credentials as environment variables:
export SRT_PHONE="010-1234-5678"
export SRT_PASSWORD="your_password"
Add these to your shell profile (~/.zshrc, ~/.bashrc, etc.) for persistence.
Security Note: Avoid committing credentials to version control.
Important: Phone number must include hyphens in the format 010-XXXX-XXXX
Usage
User-Invocable Command
Use the /srt slash command in OpenClaw:
/srt search --departure "수서" --arrival "부산" --date "20260217" --time "140000"
/srt reserve --train-id "1"
/srt reserve --retry --timeout-minutes 60
/srt reserve --retry --train-id "1,3,5" --timeout-minutes 60
/srt log -n 30
/srt list
/srt cancel --reservation-id "RES123456"
Natural Language (AI-Orchestrated)
The AI can invoke this skill based on user intent:
Examples:
- "2월 17일에 수서에서 부산 가는 기차 검색해줘" (Search trains)
- "제일 빠른걸로 예약해줘" (Reserve first available)
- "내 예약 확인해줘" (List bookings)
- "부산 예약 취소해줘" (Cancel booking)
Direct CLI Usage
# Search trains
uv run --with SRTrain python3 scripts/srt_cli.py search \
--departure "수서" \
--arrival "부산" \
--date "20260217" \
--time "140000" \
--passengers "adult=2"
# Make reservation (single attempt)
uv run --with SRTrain python3 scripts/srt_cli.py reserve --train-id "1"
# Make reservation with automatic retry - all trains (background mode recommended)
uv run --with SRTrain python3 scripts/srt_cli.py reserve --retry \
--timeout-minutes 60 \
--wait-seconds 10
# Make reservation with automatic retry - specific trains only
uv run --with SRTrain python3 scripts/srt_cli.py reserve --retry \
--train-id "1,3,5" \
--timeout-minutes 60 \
--wait-seconds 10
# Check reservation log
uv run --with SRTrain python3 scripts/srt_cli.py log -n 30
# View bookings
uv run --with SRTrain python3 scripts/srt_cli.py list --format json
# Cancel booking
uv run --with SRTrain python3 scripts/srt_cli.py cancel \
--reservation-id "RES123456" \
--confirm
Common Korean Station Names
Main SRT Stations:
- 수서 (Suseo) - Seoul SRT station
- 부산 (Busan)
- 동대구 (Dongdaegu) - Daegu
- 대전 (Daejeon)
- 천안아산 (Cheonan-Asan)
- 오송 (Osong)
- 광주송정 (Gwangju-Songjeong)
- 울산 (Ulsan)
- 포항 (Pohang)
- 경주 (Gyeongju)
- 김천구미 (Gimcheon-Gumi)
- 익산 (Iksan)
- 전주 (Jeonju)
- 목포 (Mokpo)
- 신경주 (Singyeongju)
Important: Station names MUST be in Korean (Hangul) for the SRT API to work correctly.
Date and Time Formats
- Date: YYYYMMDD (e.g.,
20260217for February 17, 2026) - Time: HHMMSS (e.g.,
140000for 2:00 PM,093000for 9:30 AM)
Tools for AI Agent
This skill provides 5 tools for managing SRT train reservations:
1. search_trains
Search for available trains between stations.
Usage:
uv run --with SRTrain python3 scripts/srt_cli.py search \
--departure "수서" \
--arrival "부산" \
--date "20260217" \
--time "120000"
Returns: JSON array of available trains with seat availability
JSON Format:
{
"success": true,
"data": [
{
"train_id": "1",
"train_number": "301",
"train_name": "SRT301",
"departure_time": "120500",
"arrival_time": "143000",
"departure_station": "수서",
"arrival_station": "부산",
"seat_available": true,
"general_seat": "예약가능",
"special_seat": "예약가능"
}
]
}
2. make_reservation
Reserve trains with optional automatic retry support.
Usage (single attempt):
uv run --with SRTrain python3 scripts/srt_cli.py reserve --train-id "1"
Usage (with retry):
# Try all trains
uv run --with SRTrain python3 scripts/srt_cli.py reserve --retry \
--timeout-minutes 60 \
--wait-seconds 10
# Try specific trains only
uv run --with SRTrain python3 scripts/srt_cli.py reserve --retry \
--train-id "1,3,5" \
--timeout-minutes 60 \
--wait-seconds 10
Options:
--train-id: Specific train(s) to reserve (comma-separated, e.g., "1" or "1,3,5"; omit to try all trains)--retry: Enable automatic retry on failure--timeout-minutes: Maximum retry duration in minutes (default: 60)--wait-seconds: Wait time between retry attempts in seconds (default: 10)
Behavior with --retry:
- Cycles through all available trains from search results
- Waits
--wait-secondsbetween attempts (plus rate-limiting delays) - Continues until success or timeout
- Logs progress to
~/.openclaw/tmp/srt/reserve.log
Returns: Reservation details with payment deadline
JSON Format:
{
"success": true,
"data": {
"reservation_id": "RES123456",
"journey_date": "20260217",
"journey_time": "120500",
"departure": "수서",
"arrival": "부산",
"train_number": "301",
"seat_number": "3A",
"payment_required": true,
"attempts": 12
}
}
Note:
- Payment must be completed manually by user via SRT app/website
- For retry mode, run in background with exec tool and periodically check logs
3. view_bookings
List all current reservations.
Usage:
uv run --with SRTrain python3 scripts/srt_cli.py list --format json
Returns: JSON array of active reservations
JSON Format:
{
"success": true,
"data": [
{
"reservation_id": "RES123456",
"journey_date": "20260217",
"journey_time": "120500",
"departure": "수서",
"arrival": "부산",
"train_number": "301",
"seat_number": "3A",
"payment_required": true
}
]
}
4. cancel_booking
Cancel a reservation by ID.
Usage:
uv run --with SRTrain python3 scripts/srt_cli.py cancel \
--reservation-id "RES123456" \
--confirm
Returns: Cancellation confirmation
JSON Format:
{
"success": true,
"data": {
"success": true,
"reservation_id": "RES123456",
"message": "Reservation cancelled successfully"
}
}
5. check_log
Check the progress of reservation attempts (especially useful for retry mode).
Usage:
uv run --with SRTrain python3 scripts/srt_cli.py log -n 30
Returns: Last N lines of reservation log file (~/.openclaw/tmp/srt/reserve.log)
Options:
-n, --lines: Number of lines to show (default: 20)
Log Format Example:
[2026-02-03 11:00:00] INFO: === SRT 예약 시작 (재시도 모드) ===
[2026-02-03 11:00:00] INFO: 타임아웃: 60분
[2026-02-03 11:00:00] INFO: 재시도 간격: 10초
[2026-02-03 11:00:00] INFO: 대상 열차: 1,3,5 (총 3개)
[2026-02-03 11:00:05] INFO: === 시도 #1 (열차 1/3) ===
[2026-02-03 11:00:05] INFO: 🎫 예약 시도 중... (열차 301, 120500)
[2026-02-03 11:00:06] WARN: ❌ 좌석 없음 (열차 301)
[2026-02-03 11:00:06] INFO: ⏳ 10초 대기 후 재시도...
[2026-02-03 11:00:26] INFO: === 시도 #2 (열차 2/3) ===
...
[2026-02-03 11:05:00] SUCCESS: ✅ 예약 성공!
Error Handling
The skill provides clear, actionable error messages:
Common Errors:
-
AuthenticationFailed
- Invalid credentials
- Solution: Check phone number and password in config
-
NoSeatsAvailable
- Train is sold out
- Solution: Try next train or different time
- Exit code: 1 (retryable)
-
StationNotFound
- Invalid station name
- Solution: Use Korean station names from the list above
-
NoTrainsFound
- No trains for specified route/time
- Solution: Try different date or time
-
RateLimitExceeded
- Too many retry attempts (max 10 per session)
- Solution: Wait a few minutes before trying again
-
NetworkError
- Connection timeout or failure
- Solution: Check internet connection and retry
Exit Codes:
0- Success1- Retryable error (e.g., no seats available)2- Fatal error (e.g., authentication failed, invalid input)
⚠️ Rate Limiting and Account Protection
To protect your SRT account from being blocked by the server:
- Automatic delays between requests:
- Minimum 3 seconds between reservation attempts
- Minimum 5 seconds between search requests
- Maximum 10 retry attempts per session
- Exponential backoff after failures (3s → 5s → 10s → 15s → 20s → 30s)
What this means for users:
- The skill will automatically wait between requests
- You'll see waiting messages like "⏳ SRT 서버 보호를 위해 대기 중 (3초)"
- If you hit the retry limit, wait a few minutes before trying again
For AI orchestration:
- The AI should inform users about delays during retries
- Example: "재시도 중입니다. 서버 보호를 위해 3초 대기합니다..."
- After 10 failed attempts, suggest alternatives like different times or dates
Natural Language Handling
When users make requests in Korean, the AI should:
-
Extract parameters from natural language:
- Stations (must convert to Korean if given in English)
- Date (relative dates like "내일", "다음주 금요일" → YYYYMMDD)
- Time (relative times like "오후 2시", "12시 이후" → HHMMSS)
- Passenger count (default to 1 if not specified)
-
Call tools in correct sequence:
- Search before reserving
- List before canceling
- Handle retry logic with rate limiting
-
Handle errors gracefully:
- If no seats available, try next train (with delays)
- If station not found, suggest correct Korean name
- Inform user about waiting times during rate limiting
-
Confirm actions in Korean:
- "예약이 완료되었습니다" (Reservation completed)
- "3초 후 다음 열차를 시도합니다" (Trying next train in 3 seconds)
- "결제는 SRT 앱에서 완료해주세요" (Complete payment in SRT app)
Real-World Usage Scenarios
Scenario 1: Simple Reservation
User: "2월 17일에 수서에서 동대구 가는거 12시이후 제일 빠른걸로 2장 예약해줘"
AI Actions:
- Parse: departure=수서, arrival=동대구, date=20260217, time=120000, passengers=adult:2
- Search trains
- Select first available train
- Reserve train
- Confirm with payment reminder
Scenario 2: Retry Until Success
User: "매진이면 성공할때까지 반복해"
AI Actions:
- Search trains
- Start background retry:
exec reserve --retry --timeout-minutes 60 --wait-seconds 10 (in background) - Create monitoring cron job (isolated session + agentTurn):
cron add --job '{ "schedule": {"kind": "every", "everyMs": 120000}, "payload": { "kind": "agentTurn", "message": "Check SRT retry log and report progress", "deliver": true, "channel": "discord" }, "sessionTarget": "isolated", "enabled": true }' cron wake --mode "now" - Inform user: "백그라운드 재시도 시작. 2분마다 업데이트합니다."
⚠️ Background Retry Monitoring
MANDATORY: Use isolated session + agentTurn cron jobs for monitoring reserve --retry.
Setup
# 1. Start background retry
exec reserve --retry --timeout-minutes 60 (background)
# 2. Create monitoring cron (isolated + agentTurn)
cron add --job '{
"schedule": {"kind": "every", "everyMs": 120000},
"payload": {
"kind": "agentTurn",
"message": "Check `srt_cli.py log -n 30`, parse progress, report to user. Delete cron if done.",
"deliver": true,
"channel": "discord"
},
"sessionTarget": "isolated",
"enabled": true
}'
# 3. Wake immediately
cron wake --mode "now"
Requirements
- sessionTarget:
"isolated"(NOT "main") - payload.kind:
"agentTurn"(NOT "systemEvent") - payload.deliver:
true - Schedule: Every 1-3 minutes depending on task duration
- Wake: Always call
cron wake --mode "now"after creating job
Why
systemEventin main session doesn't trigger agent actionagentTurnin isolated session actually executes and reports- Prevents blocking, message batching, and silent monitoring
Scenario 3: Check and Cancel
User: "내 예약 확인해주고 제일 빠른거 취소해줘"
AI Actions:
- List reservations
- Parse JSON, find earliest by date/time
- Cancel reservation
- Confirm cancellation
Scenario 4: Modify Booking
User: "부산 예약 취소하고 동대구로 다시 예약해줘"
AI Actions:
- List reservations
- Find Busan reservation
- Cancel Busan reservation
- Search for trains to 동대구 (same date/time)
- Reserve new train
- Confirm both actions
Payment Notes
IMPORTANT: This skill can search and reserve trains, but cannot process payments.
After making a reservation:
- You'll receive a reservation number
- Payment must be completed via:
- SRT mobile app (iOS/Android)
- SRT website (https://etk.srail.kr)
- Check payment deadline (usually 20 minutes after reservation)
- Unpaid reservations will be automatically cancelled
Troubleshooting
"SRT 인증 정보를 찾을 수 없습니다"
- Verify
SRT_PHONEandSRT_PASSWORDenvironment variables are set - Check your shell profile (
~/.zshrc,~/.bashrc) hasexportkeyword - Example:
export SRT_PHONE="010-1234-5678"
"검색 결과를 찾을 수 없습니다"
- Run
searchcommand beforereserve - Search results are cached in
~/.openclaw/tmp/srt/last_search.pkl
"재시도 한도 초과"
- You've made 10 reservation attempts
- Wait 5-10 minutes before trying again
- Try different trains or times
Login failures
- Verify credentials are correct
- Check if SRT service is available
- Ensure phone number format includes hyphens (010-1234-5678)
Development
Testing Locally
# Install dependencies
# Install uv if not already installed
# https://docs.astral.sh/uv/getting-started/installation/
# Configure credentials
export SRT_PHONE="010-1234-5678"
export SRT_PASSWORD="your_password"
# Test commands
uv run --with SRTrain python3 scripts/srt_cli.py search --departure "수서" --arrival "부산" --date "20260203" --time "140000"
uv run --with SRTrain python3 scripts/srt_cli.py list
Publishing to ClawHub
# Authenticate
clawhub login
# Publish
clawhub publish . \
--slug srt \
--name "SRT Korean Train Service" \
--version 0.1.2 \
--tags latest
License
MIT
Support
For issues or questions:
- File an issue on GitHub
- Check SRT service status: https://etk.srail.kr
Version History
- 0.1.3 - Retry improvements and monitoring requirements
- Unified
reservecommand with--retryflag - Added
--timeout-minutesfor time-based retry limits (default: 60) - Added
--train-idsupport for comma-separated multiple trains (e.g., "1,3,5") - Changed
--wait-secondsdefault from 20 to 10 seconds - Search includes sold-out trains by default (
available_only=False) - Monitoring: Isolated session + agentTurn cron jobs required for background retry
- Unified
- 0.1.2 - Add
--allflag for sold-out trains (deprecated) - 0.1.1 - Use
uvfor dependency management- Replace venv/pip with
uv run --with SRTrain - Environment variables only for credentials (remove config file support)
- Replace venv/pip with
- 0.1.0 - Initial release
- Search trains
- Make reservations
- View bookings
- Cancel bookings
- Rate limiting protection
- AI-friendly JSON output