huawei-cloud-ecs-lifecycle

Manage the full lifecycle of Huawei Cloud ECS (Elastic Cloud Server) instances — create, list, show details, start, stop, reboot, rename, delete, reinstall or switch the operating system, and resize (change flavor) cloud servers — plus basic ECS knowledge Q&A. Covers CRUD, power-state orchestration (start/stop/reboot) with batch support, OS reinstall/change, flavor resize, pagination, and filters. Also queries the instance's linked cross-product resources (attached EVS volumes and snapshots, VPC/subnet topology, security groups and rules, bound EIPs). A Step 0 preflight check validates the KooCLI installation, credentials, and region before anything runs. All mutating operations require explicit user confirmation before execution. Triggers include: "创建云服务器", "删除云服务器", "启动云服务器", "停止云服务器", "重启云服务器", "修改云服务器名称", "查询云服务器列表", "查询云服务器详情", "重装操作系统", "切换操作系统", "变更云服务器规格", "查询云服务器关联资源", "查询挂载云硬盘", "云硬盘快照", "ECS生命周期管理", "ECS运维", "ECS关联资源", "create ecs", "delete ecs", "start ecs", "stop ecs", "reboot ecs", "reinstall os", "resize ecs", "change flavor", "ecs lifecycle", "linked resources", "elastic cloud server management", "云服务器增删改查".

Huawei Cloud ECS Lifecycle Management

Full lifecycle management for Huawei Cloud Elastic Cloud Server (ECS) — from provisioning to decommissioning, including power-state orchestration and instance queries.

Overview

This Skill manages the complete lifecycle of ECS (Elastic Cloud Server) instances on Huawei Cloud using the KooCLI ECS service (with the huaweicloudsdkecs Python SDK as fallback):

Capability	Operations	Actions
Query	ListServersDetails, ShowServer	List instances with filters/pagination; show instance detail
Linked query	ListServerVolumeAttachments, EVS ListVolumes/ShowVolume/ListSnapshots/ShowSnapshot, VPC ShowVpc/ListSubnets/ShowSubnet/ListSecurityGroups/ShowSecurityGroup/ListSecurityGroupRules, EIP ListPublicips/ShowPublicip	Query the instance's attached EVS volumes & snapshots, VPC/subnet topology, security groups & rules, and bound EIPs
Create	CreateServers	Create one or more pay-per-use cloud servers
Power state	BatchStartServers, BatchStopServers, BatchRebootServers	Start / stop / reboot (batch)
Update	BatchUpdateServersName	Rename one or more servers
OS image	ReinstallServerWithCloudInit / ReinstallServerWithoutCloudInit, ChangeServerOsWithCloudInit / ChangeServerOsWithoutCloudInit	Reinstall the operating system, or switch the instance to a different OS image
Resize	ListResizeFlavors, ResizeServer, ResizePostPaidServer, BatchResizeServers	Query the flavors a server may be resized to; change the flavor (vCPU / memory) of one or many servers
Delete	DeleteServers	Delete one or more servers (optional EIP/data volume release)
Knowledge	—	Basic ECS FAQ (flavors, billing, auth, network, disk) answered inline

Scope note: All mutating operations (create / delete / start / stop / reboot / rename / reinstall OS / change OS / resize) are executed only after explicit user confirmation. Query operations are safe to run anytime. Reinstall OS, change OS, and resize are destructive or disruptive — see references/lifecycle-change-operations.md for the full sequence, parameters, and warnings.

Prerequisites

1. KooCLI (hcloud) installed and authenticated — see references/cli-installation-guide.md. ⚠️ This is verified at runtime by the Step 0 preflight check, not merely assumed: if hcloud is missing from PATH, unauthenticated, or no region can be resolved, the run aborts before any ECS command is attempted.
2. Region — pass --cli-region=cn-north-4 (use the account's region; e.g. cn-north-4), or the active CLI profile default is used
3. IAM permissions — the authenticated account needs the ECS permissions listed in references/iam-policies.md (least privilege); OS reinstall/change and resize additionally need the change-operation permissions documented there
4. Python 3.8+ with huaweicloudsdkecs (only needed for the SDK fallback path)

Quality Reporting Environment Variables

Environment Variable	Required	Description
SKILL_QUALITY_ENDPOINT	No	上报地址，默认 https://skillsapi.developer.myhuaweicloud.com/api/quality/report
SKILL_QUALITY_GUEST_ENDPOINT	No	Guest (unauthenticated) report endpoint
SKILL_QUALITY_NAME	No	Skill name (auto-detected)
SKILL_QUALITY_DISABLE	No	Set 1 to disable reporting (local debugging)
SKILL_QUALITY_TIMEOUT	No	Report timeout in seconds (default 3)

Workflow

0. Preflight — mandatory entry check. Run the preflight script before anything else:

   bash

   bash scripts/preflight_check.sh --region={region}
   

   It checks, in order: (1) hcloud is on PATH, (2) hcloud version runs, (3) credentials are configured, (4) the region is resolvable — then performs a read-only smoke call to prove the credentials actually work. If any check fails, stop here, print the installation / authentication guidance from references/cli-installation-guide.md, and do not attempt any ECS command. Use --skip-auth-check only for a deliberately offline run. This step exists precisely so that a missing binary or an unauthenticated account fails immediately at the entry point instead of surfacing as command not found in the middle of an operation. Exit codes: 0 = environment ready, 1 = not ready (abort the workflow), 2 = invalid usage.

1. Classify the request — knowledge question (answer inline from the Overview/FAQ below and official docs) or an ECS operation.

2. Resolve the target server(s) — for instance-specific operations, get server_id from the user, or list instances first (ListServersDetails) if the ID is unknown.

3. Query operations — build the command with the requested filters and execute directly. For linked-resource queries (attached EVS volumes/snapshots, VPC/subnets, security groups, EIPs), call ShowServer first to resolve the server's vpc_id, security-group names, and NIC port_ids, then run the corresponding EVS/VPC/EIP query from Linked Resource Queries below.

4. Mutating operations — present the exact command and parameters to the user, wait for explicit confirmation, then execute. If the CLI fails (e.g. SDK/permission error), fall back to the huaweicloudsdkecs SDK equivalent.

5. Report — return the operation result (job ID, affected instance IDs, or detailed instance info) and, for state-change operations, suggest verifying with ListServersDetails.

Core Commands

All commands below use --cli-region=cn-north-4 as the example region; use the authenticated account's region instead (or rely on the CLI profile default). Required parameters are always shown; the most common optional parameters are included. Parameter names were verified against the KooCLI --help output of each ECS operation. Example names/IDs are placeholders to be replaced with real values from the user's environment.

List ECS instances

bash

hcloud ECS ListServersDetails --cli-region=cn-north-4 --limit=25 --name=my-ecs --status=ACTIVE

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region, agent fills automatically
--project_id	Yes (auto)	string	Project ID, auto-resolved from credentials
--limit	No	integer	Page size, default 25, max 1000
--offset	No	integer	Page number, >= 0 (0 == page 1), default 1
--name	No	string	Instance name, fuzzy match (e.g. --name=my-ecs)
--status	No	string	ACTIVE / BUILD / SHUTOFF / REBOOT / ERROR / ...
--flavor	No	string	Flavor ID filter
--ip / --ip_eq	No	string	IPv4 fuzzy / exact match
--server_id	No	string	Exact IDs, comma-separated (<=100), exclusive with other filters
--enterprise_project_id	No	string	EP filter; all_granted_eps = all enterprise projects
--tags / --not-tags	No	string	Tag filter
--marker	No	string	Pagination marker (last server id of previous page)

Show ECS instance detail

bash

hcloud ECS ShowServer --cli-region=cn-north-4 --server_id=6a2f0b1c-8d4e-4f3a-9b7c-1d2e3f4a5b6c

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--project_id	Yes (auto)	string	Project ID
--server_id	Yes	string	云服务器ID (UUID)

Linked Resource Queries

Read-only cross-product lookups — safe to run anytime, no confirmation needed. Resolve server_id and the derived IDs (metadata.vpc_id, security_groups[].name, NIC port_id) from ShowServer first. Full parameter tables: references/linked-resource-queries.md.

bash

# Attached EVS volumes / snapshots
hcloud ECS ListServerVolumeAttachments --cli-region=cn-north-4 --server_id={server_id}
hcloud EVS ListVolumes --cli-region=cn-north-4 --server_id={server_id}
hcloud EVS ShowVolume --cli-region=cn-north-4 --volume_id={volume_id}
hcloud EVS ListSnapshots --cli-region=cn-north-4 --volume_id={volume_id}
hcloud EVS ShowSnapshot --cli-region=cn-north-4 --snapshot_id={snapshot_id}

# VPC / subnet topology (vpc_id from ShowServer metadata.vpc_id)
# ShowVpc needs the /v3 suffix or KooCLI prints a version notice before the JSON
# addresses keys are the vpc_id, NOT subnet ids: locate the subnet by matching the
# instance private IP against the CIDR returned by ListSubnets
hcloud VPC ShowVpc/v3 --cli-region=cn-north-4 --vpc_id={vpc_id}
hcloud VPC ListSubnets --cli-region=cn-north-4 --vpc_id={vpc_id}
hcloud VPC ShowSubnet --cli-region=cn-north-4 --subnet_id={subnet_id}

# Security groups / rules (ShowServer gives names only -> map names to IDs first)
hcloud VPC ListSecurityGroups/v3 --cli-region=cn-north-4
hcloud VPC ShowSecurityGroup/v3 --cli-region=cn-north-4 --security_group_id={sg_id}
hcloud VPC ListSecurityGroupRules/v3 --cli-region=cn-north-4 --security_group_id.1={sg_id}

# Bound EIP (join key is the NIC port_id, not server_id)
hcloud EIP ListPublicips/v2 --cli-region=cn-north-4 --port_id.1={port_id}
hcloud EIP ShowPublicip/v2 --cli-region=cn-north-4 --publicip_id={eip_id}

Create ECS instance(s)

Confirmation required before execution. Creating instances incurs billing.

bash

hcloud ECS CreateServers --cli-region=cn-north-4 --server.name=my-ecs --server.flavorRef=s6.large.2 \
  --server.imageRef=9c2f377e-7f3a-4f4e-9c1a-2b3c4d5e6f7a --server.vpcid=81a7f347-8f0e-4b2a-9c4d-5e6f7a8b9c0d \
  --server.nics.1.subnet_id=04e00f59-1a2b-3c4d-5e6f-7a8b9c0d1e2f --server.root_volume.volumetype=SSD \
  --server.security_groups.1.id=41a937e1-9b8c-7d6e-5f4a-3b2c1d0e9f8a --server.availability_zone=cn-north-4a \
  --server.count=1 --server.adminPass=Huawei@123

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--project_id	Yes (auto)	string	Project ID
--server.name	Yes	string	Instance name (1-64 chars)
--server.flavorRef	Yes	string	Flavor ID (e.g. s6.large.2), query via ListFlavors
--server.imageRef	Yes	string	Image ID (UUID), query via IMS ListImages
--server.vpcid	Yes	string	VPC ID (UUID)
--server.root_volume.volumetype	Yes	string	Root disk type: SATA/SAS/SSD/GPSSD/ESSD/...
--server.nics.1.subnet_id	Yes (in practice)	string	Subnet ID under the VPC; required unless the VPC default subnet is used
--server.security_groups.1.id	Yes (in practice)	string	Security group ID; commonly required by the API's VPC configuration
--server.availability_zone	No	string	AZ (e.g. cn-north-4a)
--server.count	No	integer	Number to create (default 1)
--server.adminPass	No	string	Initial admin password (or --server.key_name for keypair)
--dry_run	No	boolean	Pre-check without actually creating

Note: Both parameters above are marked Yes (in practice). They are optional at the API schema level, but in practice omitting them fails ECS creation in most VPC configurations: --server.nics.[N].subnet_id is required unless the VPC default subnet is used, and --server.security_groups.[N].id is commonly required by the API's VPC configuration. For pay-per-use only: CreateServers with --server.extendparam.chargingMode=0; use CreatePostPaidServers for a simpler pay-per-use flow if preferred.

Delete ECS instance(s)

Confirmation required before execution. Data on disks may be lost when --delete_volume=true.

bash

hcloud ECS DeleteServers --cli-region=cn-north-4 --servers.1.id=6a2f0b1c-8d4e-4f3a-9b7c-1d2e3f4a5b6c --delete_publicip=false --delete_volume=false

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--project_id	Yes (auto)	string	Project ID
--servers.[N].id	Yes	string	Server IDs, indexed form (--servers.1.id=...), max 1000
--delete_publicip	No	boolean	Also delete bound EIP (default false)
--delete_volume	No	boolean	Also delete attached data volumes (default false)

Start ECS instance(s)

Confirmation required before execution.

bash

hcloud ECS BatchStartServers --cli-region=cn-north-4 --os-start.servers.1.id=6a2f0b1c-8d4e-4f3a-9b7c-1d2e3f4a5b6c

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--project_id	Yes (auto)	string	Project ID
--os-start.servers.[N].id	Yes	string	Server IDs, indexed form, max 1000

Stop ECS instance(s)

Confirmation required before execution. HARD stop may lose unsaved data.

bash

hcloud ECS BatchStopServers --cli-region=cn-north-4 --os-stop.servers.1.id=6a2f0b1c-8d4e-4f3a-9b7c-1d2e3f4a5b6c --os-stop.type=SOFT

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--project_id	Yes (auto)	string	Project ID
--os-stop.servers.[N].id	Yes	string	Server IDs, indexed form, max 1000
--os-stop.type	No	string	SOFT (default, graceful) or HARD (forced)

Reboot ECS instance(s)

Confirmation required before execution. HARD reboot may lose unsaved data.

bash

hcloud ECS BatchRebootServers --cli-region=cn-north-4 --reboot.servers.1.id=6a2f0b1c-8d4e-4f3a-9b7c-1d2e3f4a5b6c --reboot.type=SOFT

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--project_id	Yes (auto)	string	Project ID
--reboot.servers.[N].id	Yes	string	Server IDs, indexed form, max 1000
--reboot.type	Yes	string	SOFT (graceful) or HARD (forced)

Rename ECS instance(s)

Confirmation required before execution.

bash

hcloud ECS BatchUpdateServersName --cli-region=cn-north-4 --name=my-ecs-renamed --servers.1.id=6a2f0b1c-8d4e-4f3a-9b7c-1d2e3f4a5b6c

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--project_id	Yes (auto)	string	Project ID
--name	Yes	string	New name (<=64 chars); batch rename applies same name to all
--servers.[N].id	Yes	string	Server IDs, indexed form, max 1000
--dry_run	No	boolean	Pre-check without applying the change

Reinstall the operating system

Confirmation required before execution. Reinstalling the OS reinitializes the system disk and reboots the instance — data on the system disk is lost. Stop the server first (BatchStopServers).

bash

# Cloud-Init variant (supports user_data injection)
hcloud ECS ReinstallServerWithCloudInit --cli-region={region} --server_id={server_id} \
  --os-reinstall.adminpass={admin_pass}

# Without Cloud-Init (no metadata.user_data parameter)
hcloud ECS ReinstallServerWithoutCloudInit --cli-region={region} --server_id={server_id} \
  --os-reinstall.keyname={key_name}

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--server_id	Yes	string	Server ID (UUID)
--os-reinstall.adminpass	No	string	New admin password (8-26 chars, complexity rules)
--os-reinstall.keyname	No	string	Keypair name for login
--os-reinstall.userid	No	string	User ID that owns the server
--os-reinstall.mode	No	string	withStopServer — stop, reinstall, then start
--os-reinstall.metadata.user_data	No	string	Cloud-Init user data (WithCloudInit only)

Reinstalling does not change the image. To switch to a different OS image use ChangeServerOsWithCloudInit instead.

Change the operating system image

Confirmation required before execution. Switching the OS also reinitializes the system disk; stop the server first.

bash

hcloud ECS ChangeServerOsWithCloudInit --cli-region={region} --server_id={server_id} \
  --os-change.imageid={image_id} --os-change.adminpass={admin_pass}

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--server_id	Yes	string	Server ID (UUID)
--os-change.imageid	Yes	string	Target image ID (UUID), query via IMS ListImages
--os-change.adminpass	No	string	New admin password
--os-change.keyname	No	string	Keypair name for login
--os-change.isAutoPay	No	string	true/false — auto-pay when the OS change is billed on a yearly/monthly instance
--os-change.mode	No	string	withStopServer — stop, change, then start
--os-change.metadata.user_data	No	string	Cloud-Init user data (WithCloudInit only)

The WithoutCloudInit variant (ChangeServerOsWithoutCloudInit) has no --os-change.metadata.user_data parameter.

Resize (change flavor)

Confirmation required before execution. Resizing restarts the instance. Always run the --dry_run=true pre-check first, then confirm and re-run without it.

bash

# 1. Query the flavors this server may be resized to
hcloud ECS ListResizeFlavors --cli-region={region} --instance_uuid={server_id}

# 2. Pre-check, no change is made
hcloud ECS ResizeServer --cli-region={region} --server_id={server_id} \
  --resize.flavorRef={flavor_id} --resize.mode=withStopServer --dry_run=true

# 3. Apply after user confirmation (drop --dry_run)
hcloud ECS ResizeServer --cli-region={region} --server_id={server_id} \
  --resize.flavorRef={flavor_id} --resize.mode=withStopServer

Parameter	Required	Type	Description
--cli-region	Yes (auto)	string	Region
--server_id	Yes	string	Server ID (UUID)
--resize.flavorRef	Yes	string	Target flavor ID, must appear in ListResizeFlavors
--resize.mode	No	string	withStopServer — allows resizing a running server (auto stop/start)
--resize.extendparam.isAutoPay	No	string	true/false — auto-pay for a yearly/monthly resize
--resize.dedicated_host_id	No	string	New dedicated host ID (DeH only)
--resize.cpu_options.hw:cpu_threads	No	integer	1 = disable hyper-threading, 2 = enable
--dry_run	No	boolean	Pre-check only; no change is made

ResizePostPaidServer (pay-per-use) and BatchResizeServers (batch, --resize.servers.[N].id) are also available. They do not share one parameter set: ResizePostPaidServer accepts --dry_run and --resize.mode but not --resize.dedicated_host_id or --resize.extendparam.isAutoPay; BatchResizeServers accepts --resize.mode, --resize.cpu_options.hw:cpu_threads and --resize.promotion.* but not --dry_run. Always confirm the accepted parameters with hcloud ECS <Operation> --help.

Full parameter tables, the complete stop→change→verify sequence, and the mapping from the operation names commonly (but incorrectly) assumed — ReinstallOs, ChangeServerFleet — to the real KooCLI operations are in references/lifecycle-change-operations.md.

Parameter Confirmation

Command	Required Parameters (besides --cli-region)
ListServersDetails	none (filters optional)
ShowServer	--server_id
Linked volumes / snapshots	--server_id (then ListVolumes / ShowVolume / ListSnapshots / ShowSnapshot by --volume_id / --snapshot_id)
Linked VPC / subnet / security groups / EIP	--vpc_id, --subnet_id, --security_group_id / --security_group_id.1, --port_id.1, --publicip_id
CreateServers	--server.name, --server.flavorRef, --server.imageRef, --server.vpcid, --server.root_volume.volumetype
DeleteServers	--servers.[N].id
BatchStartServers	--os-start.servers.[N].id
BatchStopServers	--os-stop.servers.[N].id
BatchRebootServers	--reboot.servers.[N].id, --reboot.type
BatchUpdateServersName	--name, --servers.[N].id
ReinstallServerWithCloudInit / ReinstallServerWithoutCloudInit	--server_id
ChangeServerOsWithCloudInit / ChangeServerOsWithoutCloudInit	--server_id, --os-change.imageid
ListResizeFlavors	none (filters optional)
ResizeServer / ResizePostPaidServer	--server_id, --resize.flavorRef
BatchResizeServers	--resize.servers.[N].id, --resize.flavorRef

--project_id is auto-resolved by KooCLI from the authenticated credentials; --cli-region is the only parameter the agent must always supply explicitly.

Basic ECS FAQ (Knowledge Q&A)

• What is ECS? Elastic Cloud Server — Huawei Cloud's IaaS compute service. Instances can be provisioned in minutes with flexible flavors, images, disks, and networks.
• Flavors — defined by vCPUs/memory (e.g. s6.large.2 = 2 vCPU / 4 GB). Query available flavors via hcloud ECS ListFlavors --cli-region=cn-north-4.
• Billing — pay-per-use (按需) or yearly/monthly (包年/包月); spot instances (竞价) available.
• Login — keypair (recommended) or password. For password login the initial password is set at creation (adminPass / user_data).
• Status values — ACTIVE (running), SHUTOFF (stopped), BUILD (creating), REBOOT, ERROR, MIGRATING, RESIZE, SHELVED, DELETED.

Reference Documents

• references/iam-policies.md — Least-privilege IAM policies for ECS lifecycle operations
• references/cli-installation-guide.md — KooCLI installation, authentication, and the Step 0 preflight check
• references/verification-method.md — How to verify each operation
• references/lifecycle-change-operations.md — Reinstall / change OS and resize: verified operations, parameters, sequences, and the ReinstallOs / ChangeServerFleet name mapping
• references/linked-resource-queries.md — Linked EVS volume/snapshot, VPC/subnet, security-group, and EIP read-only queries with full parameter tables
• references/dataflow-diagram.md — Mermaid data flow diagram
• references/acceptance-criteria.md — Acceptance criteria for this skill

KooCLI Command Format Standard

Every KooCLI command is built as: hcloud + service name (ECS) + operation name (PascalCase) + region parameter --cli-region=<region> + parameter key-value pairs (--key=value). Only the executable examples in Core Commands are meant to be run directly.

Feature	Description	Example
Service name	KooCLI service name (detected from metadata: ecs → invoked as ECS)	ECS
Operation name	PascalCase	ListServersDetails
Region parameter	--cli-region=<value>	--cli-region=cn-north-4
Simple parameter	--key=value	--server_id=xxx
Indexed parameter	--key.N=value	--servers.1.id=xxx

Edge Cases

Scenario	Handling
Unknown server_id / empty query result	Return "no ECS instance matches" and suggest listing first
--server_id combined with other filters	server_id wins; other filters ignored (CLI behavior)
Linked query returns empty volumes/SGs/EIPs	Report "no linked resources" — the instance may have none; verify via ShowServer output
security_groups only has names (no IDs)	Map names to IDs via VPC ListSecurityGroups/v3 before ShowSecurityGroup/v3
Instance has no bound EIP	Skip the EIP query and state the instance is not bound to a public IP
Create without subnet/security group	Explain that VPC networking config is needed; confirm before executing
User asks to delete an ACTIVE server	Warn about data loss, require explicit confirmation
--limit > 1000	CLI rejects; keep <= 1000 (default 25)
Stop/reboot with unsaved work	Prefer SOFT; explain HARD data-loss risk and confirm
CLI failure (auth/permission)	Fall back to huaweicloudsdkecs SDK; report the underlying error
No region specified	Use the active CLI profile default region; if that is empty too, the Step 0 preflight fails with guidance rather than silently picking one
hcloud missing from PATH, or no usable credentials	Step 0 preflight aborts with actionable install/auth guidance — never proceed to an ECS command
Reinstall / change OS on a running server	Stop it first (BatchStopServers), or pass --os-reinstall.mode=withStopServer / --os-change.mode=withStopServer
Reinstall / change OS requested	Warn that the system disk is reinitialized and data is lost; require explicit confirmation
Resize a running server	Pass --resize.mode=withStopServer so the platform stops and restarts it
Target flavor not offered for this instance	Run ListResizeFlavors --instance_uuid={server_id} first; a flavor absent from that list is rejected
Resize before committing	Always run --dry_run=true first; it validates the request without changing the instance
Assumed operation name ReinstallOs / ChangeServerFleet	These do not exist in KooCLI; use the real operations listed in references/lifecycle-change-operations.md

Quality Reporting

This Skill integrates scripts/skill_quality_sdk.py for execution quality reporting. Every run reports trace_id, status (success/biz_fail/sys_fail/cancel), error code, cost, and masked input/output.

Integration

• Python entry point: wrap execution with quality_context:

  python

  from skill_quality_sdk import quality_context
  with quality_context(skill_name="huawei-cloud-ecs-lifecycle") as q:
      q.input = {"operation": "ListServersDetails", "region": "cn-north-4"}
      result = run_ecs_operation()
      q.output = result
  

• CLI self-check: python3 scripts/skill_quality_sdk.py (sends one test report).

Error Code Convention

Prefix	Category	Examples
U	User input	U01 missing param, U03 no data
C	Config	C01 missing AK/SK
N	Network	N01 timeout
B	Code bug	B01 null pointer
P	Platform	P01 scheduler error

Non-blocking, fail-silent. Disable with SKILL_QUALITY_DISABLE=1 for local testing.