| name | troubleshooting-guide-authoring |
| description | Turn errors, logs, bug reports, support notes, and fixes into troubleshooting or knowledge-base articles. Distinguishes confirmed causes from possible causes and prefers safe diagnostic steps. |
Troubleshooting Guide Authoring
Convert errors, logs, and fixes into KB-quality troubleshooting articles. Distinguish confirmed causes from possible causes. Safe diagnostics first.
Use when
- producing a troubleshooting or KB article from a confirmed fix
- documenting how to diagnose a recurring error or symptom
- writing a runbook entry that can be executed by the intended audience
Do not use when
- the user wants a bug diagnosis
- the user wants release notes
- the user wants conceptual docs
Execution model
Workflow phases:
symptom -> likely causes -> diagnosis -> fix or workaround -> verification -> prevention
Symptom-first ordering. Readers arrive with a symptom, not a cause. Lead with how to recognize the symptom unambiguously.
Confirmed vs possible causes. Mark each cause confirmed (proven by source material) or possible (consistent with symptom but not proven). Do not promote without evidence.
Safe diagnostics. Diagnostic steps should be safe to run on the affected system: read-only first, then non-destructive checks, then any action that mutates state. Call out destructive steps explicitly.
Audience-executable. Steps must be runnable by the intended audience. If they require operator access, say so; if customer-runnable, keep commands minimal and explained.
Hard rules
- Distinguish
confirmed causes from possible causes for every cause listed.
- Prefer safe, read-only diagnostic steps before any mutating action.
- Mark destructive steps clearly; include rollback or recovery notes.
- Do not invent error codes, log lines, or commands.
- Keep steps executable by the intended audience; do not assume access the audience does not have.
- Include verification steps so the reader can confirm the issue is resolved.
- Add a prevention section when source material supports it.
- Works standalone; does not require any companion prompt.
Artifact format
Use only sections that matter.
symptom: how the user recognizes the problem; concrete error messages or behaviors.
likely-cause: list of causes; each marked confirmed or possible with source tie.
diagnosis-steps: ordered, safe-first steps to narrow down the cause; call out destructive steps.
fix-or-workaround: per cause, the fix or workaround; required permissions.
verification: how to confirm the issue is resolved; expected output or state.
prevention: configuration, monitoring, or operational changes that reduce recurrence.
escalation-notes: when to escalate, what information to collect first.
Response format
End each response with: scope, confidence (high|medium|low), status, key assumptions or unverified causes, next-step (only if useful).
