inspect-infrastructure-health
Skill18 starsChanged 2 months agoApache-2.0View on GitHub
Inspect reviewed Grafana infrastructure evidence for host or fleet CPU usage, memory load, disk space, and system load. Use for server-health checks, capacity symptoms, resource saturation, disk pressure, cloud or local host comparisons, and bounded infrastructure triage.
In this repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- create-fusion-deliverablesSkill
- edit-fusion-designSkill
- handoff-fusion-engineeringSkill
- inspect-fusion-designSkill
- prepare-fusion-manufacturingSkill
- reconcile-fusion-dataSkill
- run-fusion-cloud-recipesSkill
- setup-autodesk-fusionSkill
- analyze-log-signalsSkill
- investigate-application-performanceSkill
- monitor-iot-edgeSkill
- review-business-kpisSkill
- setup-grafana-observabilitySkill
- audit-ios-readinessSkill
- plan-react-native-portSkill
- port-react-native-sliceSkill
- verify-swiftui-paritySkill
- audit-review-eval-validitySkill
- benchmark-review-configsSkill
- normalize-review-runsSkill
- recommend-review-reference-architectureSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 3 are distinctive, and 8 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Run the fixed grafana.infrastructure pack and present its normalized evidence without modifying the monitored environment.
- Very specific: line 12
Use only grafana_profile_status, grafana_list_packs, grafana_run_pack, grafana_get_evidence, and grafana_resolve_link.
- Distinctive: line 14
Never accept or construct PromQL, SQL, URLs, headers, datasource IDs, panel JSON, variables, or raw provider requests.
- Distinctive: line 15
Never write to Grafana, a datasource, a host, an alert, or a deployment.
- Very specific: line 21
Call grafana_profile_status({profileAlias}).
- Very specific: line 22
Call grafana_list_packs({profileAlias}) and require the exact active pack ID grafana.infrastructure.
- Distinctive: line 23
Create one non-secret idempotency key for this attempt and reuse it for any exact retry.
- Very specific: line 24
Call grafana_run_pack with that key, the aliases, packId: "grafana.infrastructure", origin, current window, and optional baseline window.
- Very specific: line 26
Call grafana_get_evidence({reportRef,evidenceRef}) only for evidence references returned by this report and needed to support a material …
- Very specific: line 27
Call grafana_resolve_link({reportRef,linkRef}) only for a link reference returned by the same report when a trusted Grafana panel link …
- Very specific: line 29
Treat an issue as a rule-backed alert only when it has classification: SIGNED_RULE and kind: ALERT.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Inspect Infrastructure Health
Run the fixed grafana.infrastructure pack and present its normalized evidence without modifying the monitored environment.
Safety contract
- Use only
grafana_profile_status,grafana_list_packs,grafana_run_pack,grafana_get_evidence, andgrafana_resolve_link. - Accept only profile and scope aliases; one declared
LIVE,REPLAY, orSIMULATEDorigin; bounded current and baseline windows; and opaque references returned by these tools. - Never accept or construct PromQL, SQL, URLs, headers, datasource IDs, panel JSON, variables, or raw provider requests.
- Never write to Grafana, a datasource, a host, an alert, or a deployment. This skill cannot authorize remediation or a patch.
- Confirm the environment before using
LIVE. Do not silently switch amongLIVE,REPLAY, andSIMULATED.
Workflow
- Establish
profileAlias,scopeAlias, origin, and a current window no longer than seven days. Ask for an earlier equal-duration, non-overlapping baseline only when comparison matters. - Call
grafana_profile_status({profileAlias}). Stop onERRORor blocked readiness; disclosePARTIALreadiness before continuing. - Call
grafana_list_packs({profileAlias})and require the exact active pack IDgrafana.infrastructure. - Create one non-secret idempotency key for this attempt and reuse it for any exact retry.
- Call
grafana_run_packwith that key, the aliases,packId: "grafana.infrastructure", origin, current window, and optional baseline window. - Inspect CPU usage, memory load, disk free space, and system load in the returned signal order. Preserve returned units, severity, and direction; do not combine unlike units.
- Call
grafana_get_evidence({reportRef,evidenceRef})only for evidence references returned by this report and needed to support a material claim. - Call
grafana_resolve_link({reportRef,linkRef})only for a link reference returned by the same report when a trusted Grafana panel link helps the user verify the evidence. Never accept a URL as input.
Treat an issue as a rule-backed alert only when it has classification: SIGNED_RULE and kind: ALERT. Treat classification: DEVIATION as a deviation, not an anomaly. Do not infer root cause from saturation alone, and do not treat absent or stale telemetry as a healthy host.
Interpret report status exactly
COMPLETE: all expected evidence completed; evaluate each signal and issue rather than saying “healthy” by default.EMPTY: the bounded scope and window returned no usable evidence; this is an evidence gap, not zero usage.STALE: the evidence is too old for a current health claim; report its observed and collected times.PARTIAL: expected evidence is missing, failed, or mixed; state which conclusions remain unsupported.CONTRADICTORY: evidence conflicts; show the conflicting evidence references and do not choose a preferred value.ERROR: collection failed or was rejected; report the sanitized error and stop.
Completion format
Return scope, origin, current and baseline windows, overall status, a compact CPU/memory/disk/load table, issues with their returned classification, evidence gaps or contradictions, evidence references for every material claim, and trusted panel links only when resolved from report-owned references.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

