huawei-cloud-computing-query

Queries Huawei Cloud computing resources (ECS/BMS/IMS/AS/CCE), Covers ECS instances, flavors, keypairs, quotas, server groups, block devices, NICs, VNC console, BMS bare metal servers/flavors/quotas, IMS images/OS versions/members/quotas, AS scaling groups/configs/policies/activity logs/lifecycle hooks/warm pools/quotas, and CCE clusters/nodes/node pool details/node pool tags/cluster tags/quotas (read-only). No write operations. Cross-category: for cloud disks (EVS) use huawei-cloud-storage-query; for ECS create/start/stop/reboot/delete/resize use huawei-cloud-ecs-manage; for CCE workloads/namespaces/events use the container/cce kubectl-based skills. Use this skill when the user needs to query ECS instance details, list flavors, check BMS availability, browse images, view auto-scaling group/policy status, or inventory CCE clusters and nodes. Triggers: 弹性云服务器, ECS, 裸金属, BMS, 镜像, IMS, 弹性伸缩, AS, 伸缩组, 伸缩策略, 规格查询, CCE, 云容器引擎, 集群, 节点, 节点池, flavor, instance, image, scaling, cluster, node.

Huawei Cloud Resource Query

⚠️ Execution Method (Must Read): This skill executes queries via local Python scripts. Using hcloud, openstack, or other CLI tools or direct API calls is prohibited.

• Query scripts are located under the skill directory scripts/<service_category>/ (e.g., scripts/as/list_scaling_groups.py)
• Choose the execution entry based on your host, in this order:
  1. If the host provides a skill action=exec tool entry, use it: skill action=exec: skill://.venv/bin/python3 skill://scripts/<service>/<script>.py <params>
  2. Otherwise (host has no skill action=exec entry, or you are a plain shell agent), run the scripts directly with the skill's venv Python: .venv/bin/python3 scripts/<service>/<script>.py <params> (Windows: .venv\Scripts\python.exe). This is the supported fallback, not a violation.
• Never fall back to hcloud, openstack, curl on IAM, or hand-written API calls; this skill does not depend on those tools
• For specific script paths and parameters, see references/<service>/guide.md
• All paths are relative to the skill directory, which is the directory containing this SKILL.md

Overview

This skill is a standalone read-only query skill that queries Huawei Cloud resources, available specifications, and existing resource information by calling the Huawei Cloud Python SDK via local Python scripts.

This skill is applicable to the following scenarios:

1. Query available cloud resource specifications in a given region
2. Query available images for a specific operating system
3. Query cloud disk types and existing cloud disk information
4. Query existing resources and their key attributes
5. Query resources not created through Terraform or other IaC tools
6. Prepare real parameters for automation configuration, resource verification, or environment inventory
7. Obtain reusable information such as resource IDs, names, specifications, images, networks, and disks

This skill does NOT handle the following:

1. Creating resources
2. Modifying resources
3. Deleting resources
4. Guessing or fabricating information that was not queried

Explicitly Out of Scope (and where to go instead)

Requested operation	Supported here?	Use instead
ECS reinstall OS (ReinstallServer / ReinstallOS)	❌	No skill implements this yet. Use the Huawei Cloud ECS console, or KooCLI hcloud ECS ReinstallServerWithCloudInit --cli-region=<region> / ReinstallServerWithoutCloudInit --cli-region=<region>; this is a write operation
ECS resize / change flavor (ResizeServer)	❌	huawei-cloud-ecs-manage
ECS create / start / stop / reboot / delete	❌	huawei-cloud-ecs-manage
EVS cloud disks, snapshots, disk types, EVS quotas	❌	huawei-cloud-storage-query
CCE workloads (Deployment/StatefulSet), namespaces, events, pod logs	❌	container/cce/ kubectl-based skills (e.g. huawei-cloud-kubectl-cce-installer, huawei-cloud-cce-kubernetes-event-analyzer)
dry-run pre-check / rollback for change operations	❌	Not provided by any computing skill yet — see "Related Skills" below

This skill is read-only. If the user asks for a change operation, state the boundary above and point to the skill that can perform it; do not attempt the operation with this skill's scripts.

---

Capability Scope

This skill provides query capabilities through categorized scripts in the scripts directory, and usage instructions through categorized guides in the references directory. The capabilities provided by this skill include:

1. Query resource lists
2. Query individual resource details
3. Query selection information such as available specifications, images, and disk types
4. Query key identifiers and dependency relationships of existing resources

Covered resource categories and their script directories:

Category	Service	Script directory	Guide
ECS	Elastic Cloud Server	scripts/ecs/	references/ecs/guide.md
BMS	Bare Metal Server	scripts/bms/	references/bms/guide.md
IMS	Image Management Service	scripts/ims/	references/ims/guide.md
AS	Auto Scaling	scripts/as/	references/as/guide.md
CCE	Cloud Container Engine (clusters / nodes / node pool details / tags / quotas, read-only)	scripts/cce/	references/cce/guide.md

---

Related Skills (Cross-Category Guidance)

This skill deliberately stays within computing read-only queries. Related capabilities live in sibling skills; when a user request crosses the boundary below, name the right entry point instead of making the user search again:

User intent	Entry point
Query ECS / BMS / IMS / AS / CCE (read-only)	this skill
Query CCE cluster/node/node pool/quota/tags (read-only)	this skill (scripts/cce/)
Query CCE workloads, namespaces, events, pod logs	container/cce/ — huawei-cloud-kubectl-cce-installer, huawei-cloud-cce-kubernetes-event-analyzer, huawei-cloud-cce-log-analyzer
Query CCE cluster/node CPU & memory usage (metrics)	huawei-cloud-cce-metric-analyzer (scripts/huawei_cloud/cce_cluster_monitoring.py)
CCE add-ons (plugin) inspection	huawei-cloud-cce-cluster-management
CCE node-pool resize / scale-out (write)	huawei-cloud-cce-cluster-management
CCE cluster/node certificate read	Not readable — CCE v3 only exposes certificate create/revoke/rotate (write). Use the CCE console for certificate status; this skill will not call write APIs
Query cloud disks / EVS snapshots / disk types / EVS quotas	huawei-cloud-storage-query (also covers OBS / SFS / CBR)
Query OBS buckets / objects	huawei-cloud-storage-query
Query RDS instances / databases	huawei-cloud-rds-smart-service
Query VOD (video on demand) resources	huawei-cloud-vod-collector
Query CCI instances / namespaces	container/cci/ skills
Query SWR instances / repositories	huawei-cloud-swr-list
Query UCS clusters	container/ucs/ skills
Query DeH (Dedicated Host) / Fleet	Not covered by any skill yet — use the console or hcloud DEH --cli-region=<region>; track a new skill issue
ECS lifecycle change (create/start/stop/reboot/delete/resize)	huawei-cloud-ecs-manage
ECS OS reinstall	Not implemented by any computing skill yet — use the Huawei Cloud ECS console, or KooCLI hcloud ECS ReinstallServerWithCloudInit --cli-region=<region> / ReinstallServerWithoutCloudInit --cli-region=<region>
AS instance batch decommission (write)	Not implemented by any skill yet — AS write operations are out of scope for this read-only skill; track a new AS management skill issue
ECS security-group / VPC network diagnosis	huawei-cloud-vpc-network-diagnosis-management, huawei-cloud-sg-manage
Change-operation dry-run pre-check + rollback (proposed computing-mutate)	Not available yet; track in the repository issue for the computing skill gap

---

Usage Principles

Important: Script paths executed within this skill are all relative to the skill directory, which is the directory containing this SKILL.md

1. This skill only performs queries; it does not perform any write operations
2. Prioritize using information explicitly specified by the user, such as region, project, AZ, resource name, and resource ID
3. Query exactly one region per call. Do not silently loop over multiple regions to build a global inventory — regions are isolated and every extra region costs API calls plus quota. If the user truly needs a multi-region view, confirm the region list with them first, then query region by region and label each result block with its region
4. Query results must be based on actual API responses; do not speculate based on experience
5. Returned results should preserve key fields for subsequent reuse
6. When the result set is large, prioritize narrowing the scope by conditions such as region, name, id, status, and tag
7. If the current resource type does not have a corresponding script or guide, clearly state that it is not supported; do not return unreliable results
8. If the user does not provide necessary scope information and there are no default values in the environment, confirm the information before executing the query
9. Execute directly according to guide.md; do not read the script contents in the scripts directory
10. Cache output when it is large
11. You must execute -h to view usage before executing each script
12. Do not fabricate script names; execute according to the script names in guide.md. If a script name is not in guide.md, it means it is not supported

---

Prerequisites

Run the environment check once per session. The check result is cached for 30 minutes (.env_check_cache), so repeated calls are cheap:

• Linux / macOS: run bash scripts/check_env.sh (or skill action=exec: bash skill://scripts/check_env.sh when that entry exists)
• Windows: run powershell -ExecutionPolicy Bypass -File scripts/check_env.ps1 (or the skill action=exec equivalent)

Windows Note: Do not use && to concatenate commands (PowerShell 5.x does not support this); use semicolons if you need to change directories first

Verified OS scope: Linux / macOS / Windows 11. On Windows, ensure_env.py automatically selects the venv interpreter at .venv\Scripts\python.exe (POSIX uses .venv/bin/python3); do not hard-code the POSIX path on Windows.

The check runs Python >= 3.6 → install dependencies → verify SDK → verify credentials → verify service availability. A cache hit prints 命中环境检查缓存 and exits immediately; force a full re-check with --force / HW_ENV_CHECK_FORCE=1, or tune the TTL with --ttl <seconds> / HW_ENV_CHECK_TTL. If the check fails, fix the issue before running query scripts.

Environment variables (all values are secrets — never echo them):

• HW_ACCESS_KEY / HW_SECRET_KEY — required AK/SK
• HW_REGION_NAME — optional, default cn-north-4; also the fallback for every script's --region
• HW_PROJECT_ID — optional project ID; honored by AS scripts, other scripts resolve it via the IAM API when --project_id is omitted
• HW_SECURITY_TOKEN — optional, required only for temporary AK/SK

---

Execution Flow

When this skill is invoked, you must follow these steps; do not wait for the user to prompt again:

Step 1: Environment Preparation

Run the environment check once at the start of the session. It installs dependencies, verifies credentials and service availability, and caches a passing result for 30 minutes:

• Linux / macOS: bash scripts/check_env.sh (use skill action=exec: bash skill://scripts/check_env.sh when that entry exists on your host)
• Windows: powershell -ExecutionPolicy Bypass -File scripts/check_env.ps1 (or the skill action=exec equivalent)

A cache hit prints 命中环境检查缓存 and returns immediately — do not re-run the full check on every query. If the check fails, fix the issue as prompted and re-execute until it passes; use --force only when credentials or dependencies changed.

Step 2: Identify and Execute Query Scripts

1. Based on the user's query intent, read references/<service>/guide.md to determine the script path and parameters to execute
2. First execute -h to view script usage:
   • Linux / macOS: .venv/bin/python3 scripts/<service>/<script>.py -h (prefix with skill action=exec: and skill:// paths when that entry exists)
   • Windows: .venv/Scripts/python.exe scripts/<service>/<script>.py -h
3. Assemble parameters based on user requirements and execute the script:
   • Linux / macOS: .venv/bin/python3 scripts/<service>/<script>.py <parameters>
   • Windows: .venv/Scripts/python.exe scripts/<service>/<script>.py <parameters>
4. Format the results and return them to the user

Important:

• All scripts and environment check scripts are inside the skill package. Use skill action=exec when your host provides it; otherwise run them directly with the skill venv Python (supported fallback, see Execution Method)
• The venv is automatically created by the check_env script. On Linux/macOS, Python is located at .venv/bin/python3; on Windows, at .venv/Scripts/python.exe
• Do not use a bare system python3 to execute scripts (SDK dependencies live in the skill venv)
• Do not read script source code in the scripts directory; execute according to the instructions in guide.md
• Cache results when the output is large
• The --project_id parameter is optional for resource queries; when omitted, ECS/BMS/IMS/AS/CCE scripts resolve it via the IAM API (KeystoneListProjects) based on --region (AS scripts also honor the HW_PROJECT_ID environment variable). If auto-resolution fails, pass --project_id explicitly. AS API-version scripts (list_api_versions.py, show_api_version.py) call a public endpoint and do not use project_id

---

Directory Structure

The directory conventions are as follows (all paths are relative to the skill directory):

1. scripts/<resource_category>/ contains the Python query scripts for the corresponding resources. You do not need to read the script contents; execute the scripts according to the usage instructions in guide.md
2. references/<resource_category>/guide.md contains the usage guide for the corresponding resources
3. Each script is responsible for a single, clear query action
4. Each resource category maintains at least one guide.md to describe script capabilities, parameters, and usage

---

Parameter Confirmation

Before executing query scripts, confirm the following parameters:

Parameter	Required	Description
region	No	Huawei Cloud region, e.g., cn-north-4. Every script accepts --region and falls back to HW_REGION_NAME (default cn-north-4) when omitted
--project_id	No	Project ID; resolved via the IAM API (KeystoneListProjects) when omitted. Pass it explicitly if auto-resolution fails. Not used by AS API-version queries
--availability_zone	No	Availability zone; required for some resource queries

For script-specific parameters, see references/<service>/guide.md.

---

Output Format

Query results are output in JSON format, containing the following common fields:

• total: Total number of matching resources
• items: Resource list, each resource containing key fields such as id, name, and status
• ECS instance lists (list_servers_details.py, list_cloud_servers.py) include charging_mode (0→按需, 1→包周期, 2→竞价; empty renders -), and flavor details flavor_id / flavor_name / vcpus / memory(GiB) / disk / gpus
• ECS --with_related (alias --with-related) adds the whole-machine related view: fixed_ips, floating_ips, security_groups, volumes (single-server show_server.py --with_related prints them as a 关联资源 section)
• CCE list_nodes.py --all auto-paginates all nodes; --verify cross-checks the total against node-pool current_node
• CCE scripts print contextual troubleshooting hints on empty results and map 401/403/404/timeout/network errors to Chinese guidance
• Specific fields vary by resource type; see individual guide.md files for details

---

Presentation

Use these conventions when turning query output into an answer for the user:

1. Answer first, details after. Start with one sentence that answers the question (count / ID / status), then list the supporting fields.
2. Prose and bullets are the default. Use a table only when comparing a small set of comparable records (≤ 6 columns) or documenting a parameter matrix.
3. Never paste wide tables. If a script emits more than 6 columns, summarize the key fields in a bullet list (id, name, status, flavor, IP) and keep the rest available on request.
4. Bound list results. For long inventories, show the first 10–20 rows plus the total and the exact filter flags that narrow the scope; never dump every row.
5. At most 3 tables per answer. More than that reads as raw output, not an answer.
6. Empty results are prose. State the region/resource scope and the next diagnostic step in one short paragraph instead of rendering an empty table.
7. Keep secrets out. Never print AK/SK/token values, and never inline a full environment-variable dump.

---

Verification Method

1. Execute the environment check script to confirm dependencies and credentials are available
2. Use the -h parameter to view script usage and confirm parameters are correct
3. Execute a query on a known resource and compare with console data to verify result accuracy
4. Check whether the returned total count is reasonable

---

Best Practices

1. Narrow the query scope first (specify region, availability zone, etc.) to avoid returning too much data
2. Use --help to view the complete list of parameters supported by a script
3. Cache query results locally when they are large to avoid repeated requests
4. When multiple resource information is needed, query in dependency order (e.g., query VPC first, then subnet)
5. When script execution fails, check environment variables and network connectivity first

---

Reference Documentation

• Query script usage guides for each service:
  • references/ecs/guide.md — ECS instances, flavors, keypairs, quotas, server groups, block devices, NICs, VNC console
  • references/bms/guide.md — BMS bare metal servers, flavors, quotas
  • references/ims/guide.md — IMS images, OS versions, members, quotas
  • references/as/guide.md — AS scaling groups, configs, policies, activity logs, lifecycle hooks, warm pools, quotas
  • references/cce/guide.md — CCE clusters, nodes, node pools, node pool details, cluster/node-pool tags, quotas (read-only)
• references/iam-policies.md — required IAM read-only permissions
• references/acceptance-criteria.md — skill test acceptance checklist (scope, execution, response, security)

---

Version & Release Channel

• Skill version: 1.1.0 (see the version field in the frontmatter)
• Canonical source of truth: developer-skill/huaweicloud-skills on master. This directory (skills/computing/ops/huawei-cloud-computing-query/) is the maintained copy.
• Older publishing mirrors (for example the developer-skill/developer-skill copy, last synced 2026-05-29) may lag behind and can be missing CCE scripts/references. Do not treat a mirror that lacks capability as evidence that this skill lacks it; sync/mirror updates are tracked outside this skill.

---

Notes

1. This skill only provides read-only query capabilities; it does not perform any write operations
2. Do not output values of environment variables such as HW_ACCESS_KEY and HW_SECRET_KEY
3. Execute scripts through skill action=exec when the host provides it, otherwise run them with the skill venv Python (see Execution Method); never substitute hcloud/curl/API calls
4. Do not fabricate script names; strictly execute according to the names in guide.md
5. Run the environment check once per session before querying (its passing result is cached for 30 minutes)
6. When using temporary AK/SK, you need to set HW_SECURITY_TOKEN