← 返回 Skills 市场
2019
总下载
4
收藏
24
当前安装
1
版本数
在 OpenClaw 中安装
/install grafana
功能描述
Avoid common Grafana mistakes — query pitfalls, variable templating, alerting traps, and provisioning gotchas.
使用说明 (SKILL.md)
Variables and Templating
- Multi-value variable needs
$__allin regex — or only first value used ${var:csv}for comma-separated —${var:pipe}for pipe-separated in regex- Variable in query:
$varor${var}— different escaping per data source $__intervalauto-adjusts to time range — use for aggregation window- Chained variables: child depends on parent — set "Refresh" to "On time range change"
Prometheus Queries
rate()needs range vector —rate(requests_total[5m])notrate(requests_total)rate()for counters,deriv()for gauges — rate handles counter resets$__rate_intervalover hardcoded — adapts to scrape interval and dashboard range- Labels in legend:
{{label}}— multiple:{{instance}} - {{job}} - Regex filter:
metric{label=~"val1|val2"}—!~for negative match
Panel Configuration
- "No data" vs "null" are different — configure in display options
- Thresholds work on last value — not all values in range
- Min/max must match your data range — auto-scaling can hide anomalies
- Time series for trends, stat for current value — choose visualization wisely
Alerting
- Alert evaluates on server — not browser, query must work without variables
- Variables not supported in alerts — hardcode values or use templates
- Multiple conditions: AND is default — configure for OR if needed
- Alert state "Pending" before "Firing" — for duration, prevents flapping
- Notification channel must be configured — alert without channel = no notification
Dashboard Provisioning
- JSON export includes data source UID — will fail if different on import
- Use data source variables —
${DS_PROMETHEUS}substituted at runtime - Provisioned dashboards read-only by default —
allowEditing: truein provisioning - Folder must exist before dashboard provisioning — or import fails silently
Data Sources
- "Server" mode proxies through Grafana — hides credentials from browser
- "Browser" mode direct from browser — faster but exposes URL/auth
- Test connection catches most issues — but not query-specific problems
- TLS skip verify for self-signed — but fix proper certs for production
Transformations
- Order matters — transformations apply in sequence
- Outer join for combining queries — match on time or label
- Reduce for aggregating time series — last, mean, max, etc.
- Add field from calculation — combine metrics client-side
Common Mistakes
- Time range selector affects variable queries — unexpected results with "All time"
- Dashboard saved but datasource not — works locally, breaks on import
- Alert rule in dashboard vs Grafana alerting — different systems, don't mix
- Panel queries run on every refresh — high-cardinality queries slow dashboard
- Annotation queries add DB load — use sparingly on busy dashboards
安全使用建议
This skill is a read-only collection of Grafana best practices and poses low technical risk: it requires no credentials, installs, or access to your system. Two practical considerations before installing: (1) provenance — the registry entry lists an unknown source and no homepage, so verify the author or cross-check the recommendations against official Grafana docs to ensure accuracy and currency; (2) feature expectation — this skill only provides guidance (not automated provisioning or API actions). If you want an agent that actually connects to or modifies Grafana, look for a skill that explicitly requests the appropriate datasource/API credentials and has a trustworthy source.
功能分析
Type: OpenClaw Skill
Name: grafana
Version: 1.0.0
The skill bundle contains standard metadata in `_meta.json` and a `SKILL.md` file providing informational tips and best practices for using Grafana. The `SKILL.md` content is purely educational and does not contain any instructions for the AI agent to perform actions, execute commands, access files, make network calls, or manipulate its own behavior. There is no evidence of prompt injection attempts, data exfiltration, malicious execution, or other high-risk behaviors.
能力评估
Purpose & Capability
Name/description promise Grafana tips and the SKILL.md contains only Grafana-related guidance (queries, templating, alerting, provisioning, transforms). No unrelated binaries, env vars, or config paths are required.
Instruction Scope
The SKILL.md is a static set of recommendations and does not instruct the agent to read local files, access environment variables, run shell commands, or transmit data to external endpoints. It stays within the stated scope of providing advice about Grafana.
Install Mechanism
There is no install spec and no code files; nothing is written to disk or downloaded. This is the lowest-risk install model (instruction-only).
Credentials
The skill declares no required environment variables, no credentials, and no config paths. That is proportionate for a documentation/advice skill.
Persistence & Privilege
Flags are default (always: false, user-invocable, model invocation allowed). The skill does not request elevated persistence or modify other skills; its privileges are appropriate for an advice-only skill.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install grafana - 安装完成后,直接呼叫该 Skill 的名称或使用
/grafana触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release
元数据
常见问题
Grafana 是什么?
Avoid common Grafana mistakes — query pitfalls, variable templating, alerting traps, and provisioning gotchas. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 2019 次。
如何安装 Grafana?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install grafana」即可一键安装,无需额外配置。
Grafana 是免费的吗?
是的,Grafana 完全免费(开源免费),可自由下载、安装和使用。
Grafana 支持哪些平台?
Grafana 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(linux, darwin, win32)。
谁开发了 Grafana?
由 Iván(@ivangdavila)开发并维护,当前版本 v1.0.0。
推荐 Skills