Diagnostic Investigation Plan

Back to prompts
Diagnostic Investigation Plan
researchintermediateClaudeGPTGemini
troubleshooting/diagnostics/root-cause-analysis

Generate a structured research plan for a bug/error. Ranks root cause hypotheses, suggests verification steps, and flags missing info instead of guessing.

Prompt
Act as a senior systems researcher. Analyze the provided issue to create a structured investigation plan.

Inputs:
- Symptoms/Errors: {{symptoms_and_errors}}
- Relevant Context/Code: {{code_or_config_snippets}}
- Environment Details: {{os_runtime_versions_infrastructure}}

Output a numbered list with exactly these sections:

1. Core Failure Statement: Summarize the primary symptom in one concise sentence.
2. Ranked Hypotheses: List top 3 likely root causes ordered by probability. For each, cite specific evidence from the inputs that supports the link.
3. Verification Actions: For each hypothesis, provide 1-2 concrete, non-destructive steps to test validity (e.g., specific log grep patterns, config toggles, or isolation commands).
4. Confidence & Gaps: Rate your confidence (Low/Medium/High) and explicitly list any critical information missing that prevents a higher rating. Do not assume facts not present.
5. Targeted Search Queries: Generate 3 precise search strings combining tool names, error signatures, and keywords to find relevant issue trackers or forum discussions.

Constraints:
- Base all analysis strictly on provided inputs. Flag any inference as speculative.
- Never invent error codes, APIs, or behaviors not mentioned or implied by the context.
- If inputs lack sufficient detail to form a hypothesis, stop and request the missing data rather than hallucinating a solution.

When to use this#

Reach for this when debugging a complex error, intermittent failure, or unexpected behavior where the cause isn’t immediately clear. Paste your symptoms, logs, and context to receive a prioritized list of hypotheses and actionable verification steps, helping you systematically narrow down the problem before searching documentation or asking for help.

Tips#

  • Include exact error messages, stack traces, and version numbers; paraphrasing symptoms reduces the model’s ability to identify specific known issues.
  • After receiving the plan, run the verification steps and feed the results back to the model in a follow-up to eliminate hypotheses and refine the next steps.
Table of Contents