Klimkit Diagnose

Diagnose before fixing. The output should make the failure, root cause, fix, and regression proof inspectable.

Workflow

1. Ensure operator-scoped Klimkit context exists. If .klimkit/<operator>/ is missing or ambiguous, use klimkit-setup first.
2. Define the symptom in one sentence and identify the smallest observable failure signal.
3. Read relevant instructions, prior task notes, logs, and nearby tests before changing code.
4. Reproduce the failure with the cheapest deterministic command, UI path, log bundle, or fixture.
5. Minimize the failing surface: isolate the module, route, command, input, environment, or data dependency.
6. Form hypotheses from evidence. Do not patch based only on plausible explanations.
7. Instrument only when needed, and remove temporary instrumentation before handoff unless it becomes useful production diagnostics.
8. Write or update a regression test when the repo has an appropriate test surface.
9. Make the smallest robust fix.
10. Rerun the failing signal, the new regression coverage, and any blast-radius checks.
11. Record proof under .klimkit/<operator>/tasks/<feature>/: reproduction, root cause, changed files, checks run, and remaining risk.

Evidence Rules