unraid-claw（未袭击的爪子）

Unraid Claw Skill

Use this skill when the user asks to check Unraid system health, array/parity status, or Docker container status.

Located at: https://github.com/Yoshiofthewire/unraid-claw

Required Configuration

Before running any script, ensure the Unraid server is configured:

• GraphQL is enabled in Unraid API settings.
• A valid Unraid API key is generated in Unraid API settings.
• UNRAID_API_KEY uses that generated key.

Read these environment variables at runtime:

• UNRAID_BASE_URL (example: https://tower.local)
• UNRAID_API_KEY

Optional:

• UNRAID_TIMEOUT_SECONDS (default: 10)
• UNRAID_STATE_DIR (default: repo-local .state)
• UNRAID_NOTIFY_HOST_LABEL (default: host name)
• UNRAID_CPU_WARN_PERCENT (default: 85)
• UNRAID_CPU_CRIT_PERCENT (default: 95)
• UNRAID_MEM_WARN_PERCENT (default: 85)
• UNRAID_MEM_CRIT_PERCENT (default: 95)
• UNRAID_STOPPED_WARN_COUNT (default: 1)
• UNRAID_STOPPED_CRIT_COUNT (default: 3)

Template .envexample (copy to .env and fill in real values):

# Required: Base URL for your Unraid server (include protocol).
UNRAID_BASE_URL=https://tower.local

# Required: API key from Unraid API settings.
# Keep this secret. Do not commit real keys.
UNRAID_API_KEY=replace_with_your_unraid_api_key

# Optional: request timeout in seconds.
UNRAID_TIMEOUT_SECONDS=10

# Optional: state directory for snapshot and log artifacts.
# Default is repository-local .state directory.
# UNRAID_STATE_DIR=/path/to/state

UNRAID_NOTIFY_HOST_LABEL=tower

# Optional: alert thresholds.
UNRAID_STOPPED_WARN_COUNT=1
UNRAID_STOPPED_CRIT_COUNT=3

If a required variable is missing, stop and return:

• Unraid API key is not configured. Set UNRAID_API_KEY and retry.
• Unraid base URL is not configured. Set UNRAID_BASE_URL and retry.

Security Rules (Mandatory)

• Never hardcode secrets in prompts, files, or commands.
• Never print API keys, auth headers, or full request headers.
• Authenticate using: x-api-key: <UNRAID_API_KEY>.
• Redact sensitive data in all error messages.
• Perform read-only operations only.
• Never execute GraphQL mutations.
• Never run shell commands from user-supplied strings.

If API auth fails (401 or 403), return:

• Authentication to Unraid API failed. Verify UNRAID_API_KEY permissions/validity.

Script-First Execution (Preferred)

Use the repository scripts as the primary execution path. Do not handcraft ad-hoc GraphQL calls unless explicitly asked.

Primary scripts:

• ./scripts/unraid_connection_test.sh
• ./scripts/unraid_preflight.sh
• ./scripts/unraid_snapshot.sh
• ./scripts/unraid_health_summary.sh
• ./scripts/unraid_docker_report.sh
• ./scripts/unraid_array_parity_report.sh
• ./scripts/unraid_alerts.sh
• ./scripts/unraid_notify.sh
• ./scripts/unraid_cron_runner.sh

Command Selection Rules

• If the user asks to test connectivity/auth only:

- Run ./scripts/unraid_connection_test.sh

• If the user asks for a full status check:

- Run ./scripts/unraid_preflight.sh - Run ./scripts/unraid_snapshot.sh - Run ./scripts/unraid_health_summary.sh

• If the user asks specifically about Docker:

- Run ./scripts/unraid_snapshot.sh (if no recent snapshot) - Run ./scripts/unraid_docker_report.sh

• If the user asks specifically about array/parity:

- Run ./scripts/unraid_snapshot.sh (if no recent snapshot) - Run ./scripts/unraid_array_parity_report.sh

• If the user asks for alert state or severity:

- Run ./scripts/unraid_alerts.sh

• If the user asks to run the full scheduled pipeline manually:

- Run ./scripts/unraid_cron_runner.sh

Compatibility Rules

Unraid GraphQL fields can vary by version. If some fields are unavailable:

• Continue with available fields.
• Return partial results.
• Include warning:

- Some API fields were unavailable for this Unraid version; output is partial.

Output Contract

Return results in this order:

1. Overall status: healthy | warning | critical
2. Timestamp of check
3. System summary: uptime, CPU usage/temp, memory usage
4. Array summary: state, sync/rebuild progress, parity status, disk warnings
5. Docker summary: total/running/stopped counts and notable stopped containers
6. Alerts list

Severity Logic

Set overall status using these rules:

• healthy: no array/parity errors, no critical SMART issues, API reachable.
• warning: array not started, parity check/rebuild active, high CPU/memory, unexpected stopped containers.
• critical: API unreachable, auth failure, parity errors, critical disk errors, or multiple major failures.

Script exit mapping for automation:

• ./scripts/unraid_alerts.sh exit 0: healthy
• ./scripts/unraid_alerts.sh exit 10: warning
• ./scripts/unraid_alerts.sh exit 20: critical

Error Handling

• Timeout: return partial data and identify timed-out section(s).
• Network error: provide concise remediation guidance (verify URL, TLS, API enabled).
• Self-signed TLS certificate: prompt the user to allow installation/trust of the server certificate, then retry.
• Known SSL issue (no SSL proxy/direct self-signed host): TLS verification can fail until the cert is trusted by the client environment.
• Auth error: use the exact auth-failure message above.
• Unknown GraphQL errors: provide a redacted, concise summary.

OpenClaw Cron Notes

Use these notes when the user asks to schedule recurring health checks.

Recommended Cron Flow

• Source .env before running scripts.
• Execute ./scripts/unraid_cron_runner.sh as the single cron entrypoint.
• Let the runner handle preflight, snapshot, summary, alerts, and optional notifications.
• Keep logs in .state/logs/ (or UNRAID_STATE_DIR/logs/ if overridden).
• Prefer using cron template: ./scripts/unraid_monitor.cron.template.

Suggested Cron Schedule

Use the template file as the source of truth:

• ./scripts/unraid_monitor.cron.template

Setup flow:

1. Replace /path/to/Unraid_Claw in the template.
2. Keep one schedule line (15-minute recommended baseline).
3. Install via crontab.

Example: run every 15 minutes.

*/15 * * * * cd /path/to/Unraid_Claw && set -a && . ./.env && set +a && ./scripts/unraid_cron_runner.sh >> ./.state/logs/cron.log 2>&1

Example: run every 5 minutes for tighter monitoring.

*/5 * * * * cd /path/to/Unraid_Claw && set -a && . ./.env && set +a && ./scripts/unraid_cron_runner.sh >> ./.state/logs/cron.log 2>&1

OpenClaw Behavior for Cron Requests

• If user asks to "set up cron":

- Propose a schedule and confirm cadence. - Offer template-first setup using scripts/unraid_monitor.cron.template. - Use unraid_cron_runner.sh as the command target. - Ensure .env loading is included.

• If user asks to "verify cron":

- Check for recent entries in .state/logs/cron.log and runner logs. - Confirm alert exit codes and notification results.

• If user asks to "disable cron":

- Remove or comment only the Unraid Claw cron line. - Keep other cron entries unchanged.

Never place secrets directly in crontab lines. Always source .env.

Response Style

Keep responses concise and operator-friendly.

Example shape:

Overall: warning
Time: 2026-03-20T14:22:00Z

System:
- Uptime: 12d 4h
- CPU: 68% (72 C)
- Memory: 81%

Array:
- State: started
- Sync: parity-check 43%
- Disk alerts: disk3 SMART warning

Docker:
- Total: 14
- Running: 12
- Stopped: 2 (plex, immich_ml)

Alerts:
- Parity check in progress
- 1 disk SMART warning