diagnose

Files (2)

• SKILL.md
• hitl-loop.template.sh · scripts/

SKILL.md (first lines)

# Diagnose

A discipline for hard bugs. Skip phases only with explicit justification.

When exploring the code, use the project's domain glossary to get a clear mental model of the relevant modules and consult ADRs in the area you're investigating.

## Phase 1 — Build a feedback loop

**This is the skill.** Everything else is mechanical. If you have a fast, deterministic, agent-executable pass/fail signal for the bug, you will find the cause — bisection, hypothesis testing, and instrumentation just consume that signal. If you don't have one, no amount of time staring at the code will save you.

Invest disproportionate effort here. **Be aggressive. Be creative. Refuse to give up.**

### Ways to build one — try roughly in this order

1. **Failing test** at the seam that 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 snapshot.
4. **Headless browser script** (Playwright / Puppeteer) — control the UI, assert on DOM/console/network.
5. **Replay a captured trace.** Save a real network request / payload / event log to disk; replay it through the code path in isolation.
6. **Throwaway harness.** Stand up a minimal subset of the system (one service, mocked dependencies) that exercises the bug path with a single function call.
7. **Property / fuzz loop.** If the bug is "sometimes the output is wrong," run 1000 random inputs and search for the failure mode.
8. **Bisection harness.** If the bug appeared between two known states (commit, dataset, version), automate "start at state X, check, repeat" so you can run `git bisect run`.
9. **Differential loop.** Run the same input through old vs new version (or two configurations) and diff the outputs.
10. **Bash HITL (Human-in-the-Loop) script.** Last resort. If a human needs to click, guide them with `scripts/hitl-loop.template.sh` so the loop is still structured. Captured output is fed back to you.

Build the right feedback loop and the bug is 90% fixed.

### Iterate on the loop itself

Treat the loop as a product. After you have a loop, ask:

- Can I make it faster? (Cache setup, skip unrelated initialization, narrow test scope.)
- Can I make the signal sharper? (Assert on the specific symptom, not "didn't crash.")
- Can I make it more deterministic? (Fix time, seed RNG, isolate filesystem, freeze network.)

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

### Non-deterministic bugs

The goal is not a clean reproduction, but a **higher reproduction rate**. Run the trigger in a loop 100×, parallelize, add stress, shrink time windows, inject sleeps. A bug with 50% flakiness is debuggable; 1% is not — keep driving the rate up until it is debuggable.

### When you genuinely cannot build a loop

Stop and say so explicitly. List what you tried. Ask the user for: (a) access to the environment that reproduces the bug, (b) a captured artifact (HAR file, log dump, core dump, timestamped screen recording), or (c) permission to add temporary instrumentation in production. **Do not** proceed to hypotheses without a loop.

Do not advance to Phase 2 until you have a loop you believe in.

## Phase 2 — Reproduce

Run the loop. Observe the bug appear.

Confirm:

- [ ] The loop produces the failure mode the **user** described — not a different failure that happens to be nearby. Wrong bug = wrong fix.
- [ ] The failure is reproducible across multiple runs (or, for non-deterministic bugs, reproducible at a rate high enough to debug).
- [ ] You captured the exact symptom (error message, wrong output, slowness) so later phases can verify the fix actually resolves it.

Do not proceed until you reproduce the bug.

## Phase 3 — Hypothesize

Generate **3–5 ranked hypotheses** before testing any of them. Generating a single hypothesis anchors on the first plausible idea.

Each hypothesis must be **falsifiable**: state the prediction it makes.

> Format: "If <X> is the cause, then <changing Y> will make the bug disappear / <changing Z> will make it worse."

If you cannot state the prediction, the hypothesis is a guess — discard or refine.

**Show the ranked list to the user before testing.** They often have domain knowledge that re-ranks instantly ("we just deployed a change to #3"), or know hypotheses they've already ruled out. Cheap checkpoint, big time savings. Do not block — proceed with your ranking if the user is AFK.

## Phase 4 — Instrument

Each probe must map to a specific prediction from Phase 3. **Change one variable at a time.**

Tool preference:

1. **Debugger / REPL inspection** if the environment supports it. One breakpoint is worth ten logs.
2. **Targeted logs** at the boundaries that distinguish the hypotheses.
3. Never "log everything and grep."

**Tag every debug log** with a unique prefix, e.g., `[DEBUG-a4f2]`. Cleanup at the end becomes a simple grep. Untagged logs survive; tagged logs die.

**Performance branch.** For performance regressions, logs are usually the wrong approach. Instead: establish a baseline measurement (timing harness, `performance.now()`, profiler, query plan), then bisect. Measure first, fix later.

## Phase 5 — Fix + regression test

Write the regression test **before the fix** — but only if there is a **correct seam** for it.

A correct seam is one where the test exercises the **actual bug pattern** as it occurs at the call site. If the only available seam is too shallow (single-caller test when the bug needs multiple callers, unit test that doesn't replicate the chain that triggered the bug), a regression test there breeds false confidence.

**If no correct seam exists, that itself is the finding.** Note it. The code architecture is preventing the bug from being contained. Flag this for the next phase.

If a correct seam exists:

1. Turn the minimized reproduction into a failing test at that seam.
2. Watch it fail.
3. Apply the fix.
4. Watch it pass.
5. Re-run the Phase 1 feedback loop against the original (non-minimized) scenario.

## Phase 6 — Cleanup + post-mortem

Mandatory before declaring done:

- [ ] Original reproduction no longer reproduces (re-run the Phase 1 loop)
- [ ] Regression test passes (or seam absence is documented)
- [ ] All `[DEBUG-...]` instrumentation removed (`grep` for the prefix)
- [ ] Throwaway prototypes deleted (or moved to a clearly marked debug location)
- [ ] The hypothesis that proved correct is stated in the commit / PR message — so the next debugger learns

**Then ask: what would have prevented this bug?** If the answer involves architectural change (no adequate test seam, tangled callers, hidden coupling), hand off to the `/improve-codebase-architecture` skill with the specific details. Make the recommendation **after** the fix is in place, not before — you have more information now than when you started.

From hcb2019/claude-code-skills, MIT.

More from hcb2019/claude-code-skills