Capture reasoning that a command list leaves out
A command list can tell an engineer what to collect. It may not explain why a result matters, which routing context applies or what evidence would disprove a suspected cause. Troubleshooting context records those decision rules so an investigation can be reviewed and repeated.
For example, a growing error counter and a historical counter value answer different questions. The context can instruct the investigator to record the observation interval, compare changes and correlate them with the affected service before proposing an explanation.
Organize the context around a symptom
Write one focused procedure for a recognizable symptom. State its scope, applicable platforms, first checks, interpretation rules and stop conditions. Keep commands in the approved platform templates and use the context to explain when those templates are relevant.
| Field | What to record |
|---|---|
| Symptom and scope | Application, site, time window and affected user group. |
| Initial checks | Evidence needed to distinguish likely causes. |
| Interpretation | Expected behavior, meaningful deviations and counterexamples. |
| Platform constraints | Applicable vendor, software, device role and restrictions. |
| Escalation | Missing evidence, ambiguous results and responsible team. |
| Completion | Service validation required before closing the investigation. |
Apply platform-specific troubleshooting guidance
NetPrompt supports separate guidance for command selection, output interpretation and additional platform context. This helps distinguish how evidence is collected from how it is analyzed. Scope instructions to the platforms where they apply rather than relying on one large set of rules for every device.
Keep environment facts such as naming conventions and management routing arrangements separate from diagnostic rules. Retain the applicable blocked-command and approval controls when refining instructions. A contextual instruction should never be used to override the controls governing a task.
Example: investigate intermittent application latency
Illustrative scenario: an application is slow for one branch but works normally at another. The troubleshooting context asks for a shared time window, a representative application transaction, the affected path and evidence of loss or congestion. It also asks the investigator to compare a working location.
If interface counters are cumulative, a large total alone does not prove current loss. Collect comparable observations and check whether the change aligns with the reported latency. If network evidence remains inconclusive, record that limit and involve the application owner rather than declaring a network root cause.
Build a repeatable investigation with NetPrompt AI
Combine the symptom, Network Context and Network Troubleshooting Context in a focused investigation. Use available device and connected-tool evidence to test the candidate explanation. Ask the result to state the observations, reasoning, unresolved questions and proposed next action.
Example brief: “Investigate intermittent latency from the branch to the order application. Use our normal-path context and the approved diagnostic procedure. Compare current observations with a working location. Do not conclude that historical counters prove current loss. Summarize missing evidence and request approval for any change.”
- Start with a known symptom and a bounded service scope.
- Review applicability to the target devices and software.
- Collect evidence through approved checks.
- Challenge the leading explanation with a counterexample or comparison.
- Escalate uncertainty and validate the service after remediation.
Review the procedure after real incidents
A resolved incident can reveal a missing diagnostic step or a misleading assumption. Have the responsible engineer review the evidence before updating reusable guidance. Capture why the procedure changed and which platforms or services it applies to.
Use a known incident, a healthy case and a missing-evidence case when evaluating the revised procedure. A good workflow should explain a supported finding, avoid inventing a fault in the healthy case and make uncertainty clear when information is incomplete.
Common questions
- How does troubleshooting context differ from Network Context?
- Network Context describes the environment: roles, relationships and expected design. Troubleshooting Context describes the diagnostic approach: checks, interpretation, escalation and completion criteria.
- Is troubleshooting context the same as an automated runbook?
- No. Context guides the reasoning and procedure. A runbook defines executable steps and control flow. They can work together when the checks and tools are configured and validated.
- Will NetPrompt automatically learn every lesson from an incident?
- Use an explicit review process for reusable guidance. An incident summary should be checked by the responsible engineer before its findings become standing troubleshooting instructions.
Continue exploring
Further technical reading
External references explain general practices; they are not endorsements of NetPrompt AI.