windows-health-assessment

name	windows-health-assessment
description	Use when assessing role-aware Windows health, capacity, pending reboot, services, updates, recent errors, certificates, or workload readiness; use windows-troubleshooting for root-cause investigation.
metadata	{"portable":true,"compatible_with":["claude-code","codex"]}

Windows Health Assessment

Use when

• Produce a bounded host health readout before or after change.
• Identify capacity, service, reboot, event, update, or role warnings.
• Separate healthy, unhealthy, not applicable, unsupported, and unassessed checks.

Do not use when

• The cause is unknown and spans components; use windows-troubleshooting.
• The request authorises repair; assessment does not grant mutation.

Inputs

Fresh inventory, target role, expected workloads, lookback window, thresholds, maintenance context, and evidence path.

Platform and privilege boundary

R0. Some event, Defender, backup, replication, and role checks need elevation or modules; unavailable checks remain visible.

Workflow

1. Confirm target and role applicability.
2. Run Test-WseSystemHealth with explicit disk and event windows.
3. Add only role-specific checks supported by the platform row.
4. Record finding severity, confidence, source, owner, and next discriminating action.
5. Verify after change with the same oracle and comparable collection window.

Mutation, verification, and recovery

No repair occurs. Verification checks observed workload outcomes. Recovery is a handoff to the owning specialist with the pre-change baseline attached.

Stop conditions

Stop diagnosis when target, role, baseline, time window, or collection rights are insufficient to distinguish a failure from an unavailable check.

Capability contract and degraded mode

Read and local execution are required for observed findings. Without them, produce a check plan and label the assessment NOT_ASSESSED.

Outputs

Operation envelope, role-aware findings, recent events, pending-reboot state, unassessed list, ranked next actions, and evidence hash.

Decision rules

Observation	Result
Missing irrelevant role	Not applicable
Collector unavailable	Not assessed
Threshold breached with evidence	Finding with severity and owner
Symptom remains unexplained	Route to troubleshooting

Quality standards

Thresholds are explicit; findings do not overclaim root cause; collection does not modify the host.

Anti-patterns

• “Healthy” from uptime alone. Fix: use role checks.
• Treating event count as root cause. Fix: correlate time and symptom.
• Hiding unavailable checks. Fix: list them.
• Auto-restarting failed services. Fix: inspect dependencies and role.
• Comparing incompatible snapshots. Fix: require schema and target match.

References

• docs/research/source-synthesis.md
• engine/platform-matrix.yaml