alibabacloud-csas-device-health-monitor

v2026.09.24
aliyun/alibabacloud-aiops-skills00

Read-only health and risk monitoring for endpoint devices enrolled in Alibaba Cloud SASE (CSAS). Diagnose a device or scan a fleet for CPU or memory pressure, battery health, disk saturation, and online status. Use when users report a slow, hot, laggy, or draining endpoint, or ask about sustained CPU or memory use, battery health, disk capacity, online status, or devices with recent performance risks. It does not count inventory, assess idle devices, check client versions, or lock, recall, delete, or otherwise modify devices. Triggers: "CSAS device health", "slow endpoint", "CPU or memory pressure", "battery health", "disk saturation", "endpoint online status".

GitHub
安装命令
npx skhub add aliyun/alibabacloud-csas-device-health-monitor
Markdown
SKILL.md

CSAS Device Health Monitor

This skill performs only read-only CSAS API calls. A fleet scan may write a local resume file; it never alters CSAS devices.

Prerequisites

  • Aliyun CLI version 3.3.3 or later and jq 1.6 or later are required. Check with aliyun version and jq --version.
  • For first installation or a major CLI upgrade, use /bin/bash -c "$(curl -fsSL --connect-timeout 10 --max-time 120 https://aliyuncli.alicdn.com/setup.sh)". For routine updates on CLI 3.3.5 or later, use aliyun upgrade.
  • Before the first CSAS call, the script checks for aliyun-cli-csas and, if needed, installs it non-interactively with aliyun plugin install --name aliyun-cli-csas < /dev/null. This prepares only a local CLI dependency; it does not modify global CLI configuration or any CSAS cloud resource.
  • Verify only credential status with aliyun configure list. Do not read, print, enter, or embed credential values. If no valid profile is available, stop and have the user configure their identity outside this session.

Permissions and Confirmation

Read references/ram-policies.md before running the workflow. If a permission failure occurs, read that file, use the ram-permission-diagnose skill to request the listed permission, then pause until the user confirms it was granted.

Before a cloud call, confirm the device selector, region/profile context, lookback window, thresholds, scan limit, and any local resume path with the user.

Observability

Before the first cloud API invocation, read references/manifest.json from this skill root and use only its non-empty version. Stop on a missing or invalid manifest. Generate a fresh random 32-character lowercase hexadecimal session ID for this skill, reuse it only while this skill is active, and generate a distinct ID after switching skills. Never reuse or copy a documented ID.

Every actual cloud API command, including commands called by scripts/device-health.sh, must append exactly:

--user-agent "AlibabaCloud-Agent-Skills/alibabacloud-csas-device-health-monitor/{session-id} skill-version/{skill-version}"

The script reads the current manifest and creates the session ID itself. Do not use deprecated AI-mode configuration or an exported global user-agent variable.

Workflow

  1. Run scripts/device-health.sh --profile <profile> --region <region> list-devices --limit 20 --sort-by UpdateTime to discover candidate devices. Use this script rather than a manually assembled aliyun csas command, so every cloud call receives the required user agent.
  2. Run scripts/device-health.sh --profile <profile> --region <region> device-report --device <DeviceTag-or-hostname> --days 7 for a complete point-in-time report plus CPU and memory trends. Any exact CSAS DeviceTag is accepted, including non-UUID mobile tags; an exact hostname is resolved only when no matching tag exists.
  3. Run scripts/device-health.sh --profile <profile> --region <region> workload-trend --device <selector> --metric cpu --days 7 when only one workload needs analysis. Provide --from and --to instead of --days for an exact epoch window, and use --max-gap-seconds to control when sampled workload points no longer count as consecutive.
  4. Run scripts/device-health.sh --profile <profile> --region <region> health-scan --days 7 --limit 100 --resume-file .device-health-state/scan.json for a bounded synchronous fleet scan. Split larger fleets into caller-managed runs.

All commands emit one JSON result on stdout and progress or errors on stderr. A workload sample above its threshold is an observation; two consecutive high samples within the permitted gap are a warning. Device reports show both device_status (the CSAS terminal state) and collection_status (snapshot reporting recency); do not present either as a substitute for the other. Fleet scans place only warnings and critical findings in flagged; observations are not fleet alerts. CSAS workload points are approximately ten-minute samples, so no result claims an exact duration. Android, iOS, Harmony, stale, and otherwise incomplete telemetry is insufficient_data, not healthy. A zero battery value without valid collection-time and capacity evidence is reported as insufficient_data while preserving the static snapshot evidence. Use the user's language for explanations, but in every final health summary retain these exact bilingual metric labels: Disk (磁盘), Battery (电池), CPU, Memory (内存), and Collection status (上报状态). Always cover those five dimensions; name insufficient_data only for telemetry that is actually unavailable.

References

  • references/api-contract.md describes supported calls and output behavior.
  • references/cli-installation-guide.md provides CLI installation and upgrade steps.
  • references/state-files.md specifies the local scan resume format.
  • references/acceptance-criteria.md lists expected and prohibited behavior.
  • references/operational-limits.md covers rate limiting and partial results.
  • references/live-test-guide.md describes isolated, current-account read-only integration tests.
发现
标签

此技能尚未发布标签。

版本
最新版本元数据

版本

v2026.09.24

发布时间

2026年9月24日

分类

未分类

许可证

未指定

源路径

skills/security/csas/alibabacloud-csas-device-health-monitor

默认分支

master

最新提交

1ba18b8

Tree SHA

6ed356b