mx-auto

Use this skill when the user wants to:

• view the currently available local triggers
• execute an existing App trigger
• save or inspect the mode/config defaults used for trigger execution
• refresh or reuse the last callable trigger snapshot
• drive the workflow from another agent that prefers stable structured inputs and outputs

For complex or ambiguous requests, read references/learning-guide.md before deciding what to ask or run.

Core Rules

1. Keep the skill narrow. Support only trigger.execute.
2. Current active mode is local only.
3. Use direct Runtime HTTP APIs.
4. The trigger execution command body is:
   • { target: "trigger.execute", payload: { triggerId, input? }, wait, leaseTtlMs }
5. Ask only for missing local Runtime config or missing trigger identity.
6. Prefer triggerName in normal usage because it is easier for humans to work with.
7. If both triggerName and triggerId are present, resolve by exact cached trigger name first.
8. If the name does not resolve, fall back to explicit triggerId.
9. Refresh trigger snapshots from local Runtime when local access is available, instead of guessing trigger ids.
10. Do not broaden this skill into connector.collect or script.run.
11. Cloud files are kept only as reserved structure for later recovery. They are not active in this version.
12. Prefer stable machine-readable output when another agent needs to parse trigger lists:

• --list-format json

13. Default to auto-discovery before asking for anything:

• discover local Runtime base URL first
• when a user names a trigger, try running it with no input first
• ask only when auto-discovery fails or Runtime clearly says required input is missing

13. Use the wrapper entry point:

bash {baseDir}/scripts/run.sh ...

Persistent State

Preferences file:

$OPENCLAW_STATE_DIR/skill-state/mx-auto/preferences.json

Fallback:

$HOME/.openclaw/skill-state/mx-auto/preferences.json

Helper script:

bash {baseDir}/scripts/export_preference.sh show
bash {baseDir}/scripts/export_preference.sh check
bash {baseDir}/scripts/export_preference.sh apply-recommended
bash {baseDir}/scripts/export_preference.sh set-key defaultConnectionMode local
bash {baseDir}/scripts/export_preference.sh set-key defaultCloudDeviceId <device_id>

Persisted defaults:

• defaultConnectionMode
• defaultLocalBaseUrl
• defaultCloudBaseUrl
• defaultCloudDispatchPath
• defaultCloudCommandStatusPathTemplate
• defaultCloudCommandListPath
• defaultCloudDeviceId
• defaultCloudToken
• defaultWait
• defaultLeaseTtlMs
• defaultPollSec
• defaultTimeoutSec

Cached trigger snapshot:

• triggerSnapshot.loadedAt
• triggerSnapshot.sourceMode
• triggerSnapshot.registryPath
• triggerSnapshot.services[]

Defaults

Recommended defaults:

• defaultConnectionMode=local
• defaultWait=true
• defaultLeaseTtlMs=60000
• defaultPollSec=3
• defaultTimeoutSec=1200

Local Runtime discovery order:

1. MX_APP_RUNTIME_BASE_URL
2. RPA_RUNTIME_BASE_URL
3. stored defaultLocalBaseUrl
4. standard Runtime ports: 8877, 8878, 8879

• Use /local/status for optional probing.
• Load trigger catalogs from /trigger-services.
• Dispatch only through /local/commands/send.

Trigger Resolution

Resolution order is fixed:

1. exact cached trigger name match from --trigger-name
2. --trigger-id
3. fail with a clear message telling the user to refresh triggers locally, or provide triggerId if the name is not cached yet

Common Commands

List currently available local triggers:

bash {baseDir}/scripts/run.sh --connection-mode local --list-triggers

List currently available local triggers for agent parsing:

bash {baseDir}/scripts/run.sh --connection-mode local --list-triggers --list-format json

Refresh trigger snapshot from local Runtime:

bash {baseDir}/scripts/run.sh --connection-mode local --refresh-triggers

Run by trigger id in local mode:

bash {baseDir}/scripts/run.sh \
  --connection-mode local \
  --trigger-id trigger_xxx

Run by cached trigger name in local mode:

bash {baseDir}/scripts/run.sh \
  --connection-mode local \
  --trigger-name "小红书测试"

Run with explicit business input only when the user intentionally overrides a trigger's built-in configuration:

bash {baseDir}/scripts/run.sh \
  --connection-mode local \
  --trigger-name "小红书测试" \
  --input-json '{"keyword":"AI employee"}'

Final Reply

When a human asks for listing, prefer a short human-readable trigger list:

• trigger count
• trigger names
• status
• summary only when it is actually useful

When another agent needs the list, prefer structured JSON with:

• mode
• sourceMode
• loadedAt
• registryPath
• triggerCount
• triggers[]

When execution succeeds, summarize:

• mode
• trigger identity
• commandId when available
• final status
• the key result

When the run fails, clearly distinguish:

• local Runtime reachability or authorization problems
• trigger list refresh failure
• command execution failure

Use this minimal user-facing guidance when local auto-discovery fails:

“本地模式会先自动查找本地 Runtime。通常不需要手动配置。 如果自动发现失败，优先检查：

1. Runtime 是否正在本机运行
2. Runtime 的本地配置是否已准备好
3. 如果路径或端口不是默认值，再显式设置对应环境变量或运行参数”