SRT Korean Train Service Skill#

Prerequisites#

• Environment variables SRT_PHONE (format: 010-XXXX-XXXX) and SRT_PASSWORD must be set before running scripts.

Reference#

Environment variables:

VariableRequiredDescriptionSRT_PHONE✅SRT account phone number (hyphens required: 010-XXXX-XXXX)SRT_PASSWORD✅SRT account passwordSRT_DATA_DIRoptionalDirectory for logs, cache, and state files. Defaults to system temp dir ( /tmp/srt).

Station names (Korean only): 수서, 부산, 동대구, 대전, 천안아산, 오송, 광주송정, 울산, 포항, 경주, 김천구미, 익산, 전주, 목포, 신경주

Date: YYYYMMDD · Time: HHMMSS (e.g. 200000 = 20:00)

---

Commands#

Search Trains#

bash

cd <project_dir> && uv run --with SRTrain python3 scripts/srt_cli.py train search \
  --departure "수서" --arrival "동대구" --date "20260227" --time "200000"

Search params and results are cached (in SRT_DATA_DIR) and required by reserve.

Reserve (one-shot)#

bash

cd <project_dir> && uv run --with SRTrain python3 scripts/srt_cli.py reserve one-shot --train-id "1"

--train-id is the 1-based index from search results. Must run train search first.

View Reservations#

bash

cd <project_dir> && uv run --with SRTrain python3 scripts/srt_cli.py reserve list --format json

Cancel Reservation#

bash

cd <project_dir> && uv run --with SRTrain python3 scripts/srt_cli.py reserve cancel \
  --reservation-id "RES123456" --confirm

---

Continuous Monitoring (취소표 모니터링)#

For "keep trying until a seat opens" requests, do not loop inside a cron job. Instead: run srt_cli.py reserve retry as a persistent background process, then create a separate cron job to read the log and report.

Step 1: Search (populate cache)#

bash

cd <project_dir> && uv run --with SRTrain python3 scripts/srt_cli.py train search \
  --departure "수서" --arrival "동대구" --date "20260227" --time "200000"

Note the train_id of the target train from the results.

Step 2: Start background retry process#

bash

LOG_FILE=<choose_any_path>.log
PID_FILE=<choose_any_path>.pid
cd <project_dir> && nohup uv run --with SRTrain python3 scripts/srt_cli.py reserve retry \
  --train-id <id> --timeout-minutes 1440 --wait-seconds 10 \
  --log-file "$LOG_FILE" > /dev/null 2>&1 &
echo $! > "$PID_FILE"

The script prints LOG_FILE: <path> on startup — capture this to know exactly where logs are written. You may also set SRT_DATA_DIR to control where auto-generated logs and cache files are placed.

Path safety: SRT_DATA_DIR and --log-file are validated at runtime to resolve within the user's home directory or system temp dir only. Paths that escape these boundaries (e.g. via ../) are rejected.

reserve retry options:

OptionDefaultDescription--train-id(all)1-based index from search; comma-separated for multiple--timeout-minutes60Total duration. Use 1440 for 24h--wait-seconds10Delay between attempts--log-fileautoExplicit log file path (overrides SRT_DATA_DIR default)

Log markers to watch for:

• === 시도 #N — attempt number
• SUCCESS — reservation succeeded (contains 예약번호, 좌석)
• TIMEOUT — timed out without success

Step 3: Create periodic reporting cron job#

Create an isolated agentTurn cron job (every 15 min) with --no-deliver (delivery mode: none). The agent must use the message tool to post directly to the Discord channel — do not use --announce (the announce queue can fail with a gateway pairing error in isolated sessions).

CLI:

bash

openclaw cron add \
  --agent srt \
  --name "SRT 모니터링 보고 (15분마다)" \
  --every 15m \
  --session isolated \
  --no-deliver \
  --message "..."

Agent message must instruct:

1. Check process status: bash

   cd <project_dir> && uv run --with SRTrain python3 scripts/srt_cli.py reserve status --pid-file <pid_file>
   
   

2. Read log tail: tail -50 <log_file>

3. Summarise attempt count, last attempt time, success/failure

4. Send report via message tool ( channel=discord, target=<channel_id>)

5. On SUCCESS in log → include 예약번호/좌석 in message, then remove this cron job and termination job

6. On NOT_RUNNING without SUCCESS → report crash, remove this cron job

The message payload must include this job's own ID and the termination job ID so it can self-remove.

Step 4: Create termination job#

Create an isolated agentTurn at-schedule cron job ( --no-deliver, --delete-after-run) at the deadline.

CLI:

bash

openclaw cron add \
  --agent srt \
  --name "SRT 모니터링 종료" \
  --at "<ISO UTC time>" \
  --session isolated \
  --no-deliver \
  --delete-after-run \
  --message "..."

Agent message must instruct:

1. Stop the process: bash

   cd <project_dir> && uv run --with SRTrain python3 scripts/srt_cli.py reserve stop --pid-file <pid_file>
   
   

2. Remove the reporting cron job by ID

3. Read final log and send outcome via message tool to Discord channel

---

JSON Output#

Search result item:

json

{
  "train_number": "369",
  "departure_time": "200000",
  "arrival_time": "213600",
  "departure_station": "수서",
  "arrival_station": "동대구",
  "seat_available": false,
  "general_seat": "매진",
  "special_seat": "매진",
  "train_id": "1"
}

Reservation result:

json

{
  "success": true,
  "data": {
    "reservation_id": "RES123456",
    "train_number": "369",
    "seat_number": "3A",
    "payment_required": true
  }
}

Exit codes: 0 = success · 1 = retryable (no seats) · 2 = fatal

---

Error Handling#

ErrorCauseResolutionAuthenticationFailedWrong credentialsCheck SRT_PHONE / SRT_PASSWORD``NoSeatsAvailableSold outUse --retry or try different trainStationNotFoundInvalid nameUse Korean station names aboveNoTrainsFoundNo trains foundTry different date/timeRateLimitExceededToo many attemptsWait a few minutes

---

Natural Language Handling#

Extract from Korean input:

• Stations → Korean names (수서, 동대구, etc.)
• Date → relative ("내일", "다음주 금요일") to YYYYMMDD
• Time → ("20시 이후", "오후 2시") to HHMMSS
• Passenger count → default 1 if not specified

Patterns:

• "검색해줘" → train search
• "예약해줘" (one-shot) → train search then reserve one-shot
• "취소표 나오면 잡아줘 / 될 때까지 돌려줘" → Continuous Monitoring flow above
• "내 예약 확인해줘" → reserve list
• "취소해줘" → list then cancel

Payment Note#

Reservations must be paid via SRT app or https://etk.srail.kr within ~20 minutes of reservation.