01
Gather the investigation facts
Collect the request, known system boundaries, evidence, impact, and practical limits without trying to solve the issue yet.
Made for Engineers
Choose What to Investigate First in an Unfamiliar System
Also useful for: people in Software Engineering
20 min
Intermediate
Decisions
Get Something Done
Need to Make a Decision
Prioritize the fastest, most useful checks before changing code or committing to a solution.
We'll use them to write your prompt.
Completed ✓
Make Your Play
Give us the situation. We’ll build the prompt.
Reduce guesswork by ranking investigation steps against current evidence, uncertainty, likely impact, and effort.
Your answers save as you type.
What you'll do
This is the plan inside your prompt. Add your details first. Then follow these steps with ChatGPT or Claude.
02
Generate and rank useful checks
Use the primary prompt to compare possible checks by how much uncertainty they remove, how strongly they affect the decision, and what they cost.
03
Commit to the first investigation path
Choose the first check, confirm its stopping condition, and retain the branch for what happens after each likely result.
Make Your Play
Choose What to Investigate First in an Unfamiliar System
Tell us what the prompt needs to know
Start with the questions below. Each answer personalizes the playbook and builds the prompt you’ll use at the end.
Your situation
Select the kind of work that triggered the investigation.
Describe the reported behavior, requested change, and expected outcome in plain language.
0/10000
List the components, dependencies, data flows, owners, or environments already known.
0/10000
List logs, screenshots, reproduction notes, recent changes, tests, code references, or other clues already available.
0/10000
Describe who is affected, how serious the outcome is, and what could go wrong during investigation or changes.
0/10000
State the available time, system access, environment limits, and people who can answer questions.
0/10000
This playbook was updated. Review your inputs before copying.
What you'll do
Check a step when you finish it.
01
Gather the investigation facts
Collect the request, known system boundaries, evidence, impact, and practical limits without trying to solve the issue yet.
02
Generate and rank useful checks
Use the primary prompt to compare possible checks by how much uncertainty they remove, how strongly they affect the decision, and what they cost.
03
Commit to the first investigation path
Choose the first check, confirm its stopping condition, and retain the branch for what happens after each likely result.
Fill in the fields above to generate your prompt preview.
Assembled prompt
Beginner instructions
Treat “confirmed” as something supported by a log, reproducible behavior, code, configuration, documentation, or a reliable source. If evidence is thin, say so. Keep the first investigation step small enough to complete without redesigning the system.
Advanced instructions
Add architecture boundaries, observability gaps, dependency ownership, rollback risk, and test coverage where relevant. Challenge steps that gather information without changing a decision. Prefer probes that distinguish between multiple plausible causes at once.
Use this when
You're handling a bug or change request in a system you don't understand well, and several possible investigation paths could consume your time.
You'll need
- Request type
- Request summary
- System context
- Available evidence
- Impact and risk
- Investigation constraints
You'll get
- Confirmed facts, inferences, and unknowns clearly separated
- Investigation options ranked by evidence, uncertainty, impact, and cost
- First investigation step specific enough to start immediately
- Stopping condition defined for every recommended check
- Next step identified for each likely first finding
- Requester questions separated from technical checks.
By the end of this playbook
You will have:
- Confirmed facts, inferences, and unknowns clearly separated
- Investigation options ranked by evidence, uncertainty, impact, and cost
- First investigation step specific enough to start immediately
- Stopping condition defined for every recommended check
- Next step identified for each likely first finding
- Requester questions separated from technical checks.
Examples
Progress 1 of 3