Byted Util Volcengine Cloud Detect
Volcengine Cloud Detect (云拨测) API integration for creating detection tasks, retrieving results, g…
by bytedance · Apache-2.0
The skill
Volcengine Cloud Detect (云拨测) API integration for creating detection tasks, retrieving results, generating reports, and managing detection nodes. Use when the user wants to (1) create network detection/dialing tasks (拨测任务), (2) check task results or generate detection reports, (3) list or search detection nodes (拨测节点), (4) manage existing tasks (list/stop/restart/delete). Triggers include natural language requests about website monitoring, network latency testing, HTTP detection, ping tests, DNS tests, node queries, or report generation for Volcengine cloud detect service.
Volcengine Cloud Detect
Interact with Volcengine Cloud Detect API (云拨测) through natural language. All API calls are handled via scripts/volc_detect.py. For detailed API parameters, read references/api_docs.md.
Prerequisites
Volcengine credentials are required:
VOLC_ACCESSKEY: Access KeyVOLC_SECRETKEY: Secret Key
Before executing any command, check whether both credentials exist in environment variables:
test -n "$VOLC_ACCESSKEY" && test -n "$VOLC_SECRETKEY"
If either value is missing, ask the user to provide the missing AK/SK, then write them to the current shell environment and persist them to the user's shell profile:
export VOLC_ACCESSKEY="USER_PROVIDED_ACCESS_KEY"
export VOLC_SECRETKEY="USER_PROVIDED_SECRET_KEY"
SHELL_PROFILE="${ZDOTDIR:-$HOME}/.zshrc"
grep -q '^export VOLC_ACCESSKEY=' "$SHELL_PROFILE" 2>/dev/null \
&& sed -i '' 's|^export VOLC_ACCESSKEY=.*|export VOLC_ACCESSKEY="USER_PROVIDED_ACCESS_KEY"|' "$SHELL_PROFILE" \
|| printf '\nexport VOLC_ACCESSKEY="USER_PROVIDED_ACCESS_KEY"\n' >> "$SHELL_PROFILE"
grep -q '^export VOLC_SECRETKEY=' "$SHELL_PROFILE" 2>/dev/null \
&& sed -i '' 's|^export VOLC_SECRETKEY=.*|export VOLC_SECRETKEY="USER_PROVIDED_SECRET_KEY"|' "$SHELL_PROFILE" \
|| printf 'export VOLC_SECRETKEY="USER_PROVIDED_SECRET_KEY"\n' >> "$SHELL_PROFILE"
Never run scripts/volc_detect.py until both VOLC_ACCESSKEY and VOLC_SECRETKEY are available.
Default Parameters for CreateTask
When user creates a task without specifying all parameters, use these defaults:
| Parameter | Default Value | Notes |
|---|---|---|
| ProjectName | default |
Always use unless user specifies |
| Type | 1 (HTTP(S)) |
Task type |
| Name | {current_time}_claw_detect |
Format: YYYYMMDD_HHMMSS_claw_detect |
| Address | (required) | Must ask user if not provided |
| NodeCount | 5 |
Samples per line |
| IntervalSeconds | 60 |
Detection frequency |
| FinishTime | current_time + 90 seconds |
1.5 minutes after creation |
| LineIdList | (required) | Must obtain from ListNodes |
Task Type Mapping: 1=HTTP(S), 2=DNS, 3=PING, 5=UDP, 6=TCP, 8=Upload, 9=Download, 11=PageElement
Interval Options (seconds): 60, 120, 180, 300, 600, 900, 1200, 1800, 3600, 7200, 10800, 21600, 43200, 86400
Workflow: Creating a Detection Task
- Collect required info: Ask user for
Address(target URL). If not provided, prompt for it. - Determine node selection: Before creating task, ask user about detection requirements to filter nodes:
- Target region? (e.g., mainland China, specific provinces/cities)
- ISP preference? (e.g., China Telecom, China Mobile, China Unicom)
- Node type? (IDC = data center, LastMile = end user)
- Fetch and filter nodes: Run
list-nodes, then fuzzy-search based on user requirements. - Confirm LineIdList: Show matched nodes to user and confirm selection.
- Create task: Run
create-taskwith all parameters including confirmed LineIdList. - Return TaskId: Present the created task ID to user.
Node Selection Conversation Guide
When user says "create a detection task for baidu.com", respond by:
- Confirming the target address
- Asking about geographic/ISP requirements
- Running
list-nodesto show available options - Using
fuzzy-search-nodesto filter based on user's description
Example user requirements to node filters:
- "mainland China + Telecom" →
--is-mainland true+ search "电信" - "Beijing IDC nodes" → search "北京 IDC"
- "Shanghai and Guangzhou Mobile" → search "上海 移动" and "广州 移动"
- "all over the country" →
--is-mainland true, select diverse nodes
Workflow: Getting Results & Reports
- Get task results: Run
get-resultwith TaskId and time range (must be <= 1 hour). - Generate report: Run
generate-reportto produce summary statistics. - Present findings: Show key metrics in a concise format.
Report Format
Present reports using this structure:
## Detection Report for {TaskName} (ID: {TaskId})
**Time Range**: {start} ~ {end}
**Total Samples**: {N}
**Success Rate**: {X}%
### Latency Statistics (ms)
| Metric | Value |
|--------|-------|
| Avg | {avg} |
| Min | {min} |
| Max | {max} |
| P90 | {p90} |
### Status Code Distribution
| Code | Count |
|------|-------|
| ... | ... |
### Issues
- {N} failures from {region-isp} nodes
- Top error: {error_code} ({count} occurrences)
Workflow: Managing Tasks
- List tasks:
list-taskswith optional filters (--name, --address) - Stop task:
stop-task --task-id ID - Restart task:
restart-task --task-id ID - Delete task:
delete-task --task-id ID(confirm with user first) - Get task details:
get-task --task-id ID
Workflow: Node Discovery
- List all nodes:
list-nodes [--is-mainland true|false] - Filter by type:
list-nodes --line-types 1,2(1=LastMile, 2=IDC) - Fuzzy search:
fuzzy-search-nodes "search query"
Script Commands Reference
All commands use python scripts/volc_detect.py:
# Create task
python scripts/volc_detect.py create-task \
--address "https://example.com" \
--name "task_name" \
--type 1 \
--node-count 5 \
--interval-seconds 60 \
--finish-time TIMESTAMP \
--line-ids 228,229,230 \
--http-method 1
# Get results
python scripts/volc_detect.py get-result \
--task-id 12345 \
--start-time 1672531200 \
--end-time 1672534800 \
--page-size 500
# List nodes
python scripts/volc_detect.py list-nodes --is-mainland true
# Fuzzy search nodes
python scripts/volc_detect.py fuzzy-search-nodes "北京 电信 IDC"
# Generate report
python scripts/volc_detect.py generate-report \
--task-id 12345 \
--start-time 1672531200 \
--end-time 1672534800
# Task management
python scripts/volc_detect.py list-tasks --page-size 20
python scripts/volc_detect.py stop-task --task-id 12345
python scripts/volc_detect.py restart-task --task-id 12345
python scripts/volc_detect.py delete-task --task-id 12345
python scripts/volc_detect.py get-task --task-id 12345
Key API Notes
- Time constraints: GetTaskResult time range must be <= 1 hour
- Page size limit: GetTaskResult max PageSize is 500
- Node naming format:
国家-省份-城市-运营商(类型), e.g.,中国-北京-北京市-中国电信(IDC) - Task status: 2=Running, 4=Paused, 6=Ended
- Project scoping: All task-scoped commands (
create-task,get-result,get-task,list-tasks,stop-task,restart-task,delete-task,generate-report,list-nodes) accept an optional--project-name.create-taskdefaults todefault; the others default to the account's default project when omitted.
Resources
- scripts/volc_detect.py: Complete API client with auth, all API operations, fuzzy search, and report generation
- references/api_docs.md: Detailed API parameter reference for all 8 endpoints
Use this skill
Connect your agent to the Skill Store MCP server and ask for it by ref: bytedance/byted-util-volcengine-cloud-detect-48255960263.
Details
- Publisher
- bytedance
- Category
- Software Development › DevOps & cloud
- License
- Apache-2.0
- Updated
- 2026-10-16
References
More from bytedance
- Blocking Io Guard Ensure async-path backend code that could block the asyncio event loop is protected by a teeth-ve…
- Deerflow Maintainer Orchestrator Use when a DeerFlow maintainer needs comment-only GitHub issue or PR handling: resolve issue/PR s…
- Engineer System Change Evaluate and carry out non-trivial software-system changes from first principles.
- Smoke Test End-to-end smoke test skill for DeerFlow.