Diagnose

A discipline for hard bugs. Skip phases only when explicitly justified.

Core insight: If you have a fast, deterministic, agent-runnable pass/fail signal for the bug, you will find the cause. Everything else is mechanical. If you don't have one, no amount of staring at code will save you.

Phase 1 — Build a feedback loop

This is the skill. Spend disproportionate effort here.

Construction strategies — try roughly in this order

1. Failing test at whatever seam reaches the bug — unit, integration, e2e.
2. Curl / HTTP script against a running dev server.
3. CLI invocation with a fixture input, diffing stdout against a known-good snapshot.
4. Headless browser script (Playwright / Puppeteer) — drives the UI, asserts on DOM/console/network.
5. Replay a captured trace. Save a real request/payload/event log to disk; replay through the code path in isolation.
6. Throwaway harness. Spin up a minimal subset of the system (one service, mocked deps) exercising the bug path with a single function call.
7. Property / fuzz loop. "Sometimes wrong output" → run 1000 random inputs and look for the failure mode.
8. Bisection harness. Bug appeared between two known states → automate "boot at state X, check, repeat" so you can git bisect run it.
9. Differential loop. Same input through old-version vs new-version (or two configs), diff outputs.
10. HITL script. Last resort. If a human must click, drive them with a structured bash script so the loop is still reproducible. Captured output feeds back to you.

Iterate on the loop itself

Treat the loop as a product:

• Faster? Cache setup, skip unrelated init, narrow scope.
• Sharper signal? Assert on the specific symptom, not "didn't crash."
• More deterministic? Pin time, seed RNG, isolate filesystem, freeze network.

A 30-second flaky loop is barely better than no loop. A 2-second deterministic loop is a debugging superpower.

Non-deterministic bugs

Goal: raise reproduction rate. Loop 100x, parallelise, add stress, narrow timing windows, inject sleeps. A 50%-flake is debuggable; 1% is not.

When you genuinely cannot build a loop

Stop and say so. List what you tried. Ask the user for: (a) access to the reproducing environment, (b) a captured artifact (HAR file, log dump, core dump, screen recording), or (c) permission to add temporary production instrumentation. Do NOT proceed to hypothesise without a loop.

Phase 2 — Reproduce

Run the loop. Watch the bug appear. Confirm:

• The failure matches what the user described — not a nearby different failure.
• Reproducible across multiple runs (or high enough rate for non-deterministic bugs).
• Exact symptom captured (error message, wrong output, timing) for later verification.

Phase 3 — Hypothesise

Generate 3-5 ranked hypotheses before testing any. Single-hypothesis generation anchors on the first plausible idea.

Each hypothesis must be falsifiable:

"If \x3CX> is the cause, then \x3Cchanging Y> will make it disappear / \x3Cchanging Z> will make it worse."

If you can't state the prediction, the hypothesis is a vibe — discard or sharpen it.

Show the ranked list to the user before testing. They often re-rank instantly with domain knowledge. Don't block — proceed with your ranking if AFK.

Phase 4 — Instrument

Each probe maps to a specific prediction from Phase 3. One variable at a time.

Tool preference:

1. Debugger / REPL if available. One breakpoint beats ten logs.
2. Targeted logs at boundaries that distinguish hypotheses.
3. Never "log everything and grep."

Tag every debug log with a unique prefix: [DEBUG-a4f2]. Cleanup = single grep. Untagged logs survive; tagged logs die.

Performance bugs: logs are usually wrong. Establish a baseline measurement (timing harness, profiler, query plan), then bisect. Measure first, fix second.

Phase 5 — Fix + regression test

Write the regression test before the fix — but only if there's a correct seam.

A correct seam exercises the real bug pattern as it occurs at the call site. If the only seam is too shallow, a regression test there gives false confidence.

If no correct seam exists, that itself is the finding — note it.

1. Turn the minimised repro into a failing test at the seam.
2. Watch it fail.
3. Apply the fix.
4. Watch it pass.
5. Re-run the Phase 1 loop against the original scenario.

Phase 6 — Cleanup + post-mortem

Before declaring done:

• Original repro no longer reproduces (re-run Phase 1 loop)
• Regression test passes (or absence of seam is documented)
• All [DEBUG-...] instrumentation removed (grep the prefix)
• Throwaway harnesses deleted
• Root cause stated in the commit/PR message

Then ask: what would have prevented this bug? If the answer involves architectural change, note it for the user — don't bundle it into this fix.

Completion Status

• DONE — root cause found, fix applied, regression test written, all tests pass
• DONE_WITH_CONCERNS — fixed but cannot fully verify (intermittent, needs staging)
• BLOCKED — root cause unclear after investigation, escalated