Runs a phased hard-bug workflow: invent a fast deterministic red-capable loop, minimise the repro, rank 3-5 falsifiable hypotheses, probe one variable, then land a seam-correct regression. Trigger on diagnose requests, stubborn flakes, throws, or performance regressions. Not a product-spec chair; never start hypothesizing before that loop has already been executed once.
Runs a phased hard-bug workflow: invent a fast deterministic red-capable loop, minimise the repro, rank 3-5 falsifiable hypotheses, probe one variable, then land a seam-correct regression. Trigger on diagnose requests, stubborn flakes, throws, or performance regressions. Not a product-spec chair; never start hypothesizing before that loop has already been executed once.
A disciplined, phased workflow for hard bugs and performance regressions. Skip phases only when explicitly justified.
When to Use
Use this skill when the user says "diagnose", "debug this", or reports something broken, throwing, failing, or slow — especially when the bug is non-obvious, intermittent, or resists a quick fix.
When exploring the codebase, read CONTEXT.md (if it exists) to get a clear mental model of the relevant modules, and check ADRs in the area you're touching.
Prerequisites
Access to the codebase and ability to run tests / scripts / dev servers.
For HTTP-based bugs: a running dev server or the ability to start one.
For non-deterministic bugs: willingness to invest time raising the reproduction rate before debugging.
Optional: scripts/hitl-loop.template.sh — load this template from scripts/ when a human must be in the loop (clicking UI, manual steps). It structures the HITL cycle so captured output feeds back to the agent.
Procedure
Phase 1 — Build a Feedback Loop
This is the skill. Everything else is mechanical. If you have a tight pass/fail signal for the bug — one that goes red on this bug — you will find the cause. If you don't have one, no amount of staring at code will save you.
Spend disproportionate effort here. Be aggressive. Be creative. Refuse to give up.
Ways to construct a loop — try in roughly this order
Failing test at whatever seam reaches the bug — unit, integration, e2e.
Curl / HTTP script against a running dev server.
CLI invocation with a fixture input, diffing stdout against a known-good snapshot.
Headless browser script (Playwright / Puppeteer) — drives the UI, asserts on DOM/console/network.
Replay a captured trace. Save a real network request / payload / event log to disk; replay it through the code path in isolation.
Throwaway harness. Spin up a minimal subset of the system (one service, mocked deps) that exercises the bug code path with a single function call.
Property / fuzz loop. If the bug is "sometimes wrong output", run 1000 random inputs and look for the failure mode.
Bisection harness. If the bug appeared between two known states (commit, dataset, version), automate "boot at state X, check, repeat" so you can it.
git bisect run
Differential loop. Run the same input through old-version vs new-version (or two configs) and diff outputs.
HITL bash script. Last resort. If a human must click, drive them with scripts/hitl-loop.template.sh so the loop is still structured. Captured output feeds back to you.
Tighten the loop
Treat the loop as a product. Once you have a loop, tighten it:
Can I make it faster? (Cache setup, skip unrelated init, narrow the test scope.)
Can I make the signal sharper? (Assert on the specific symptom, not "didn't crash".)
Can I make it 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 one is tight — a debugging superpower.
Non-deterministic bugs
The goal is not a clean repro but a higher reproduction rate. Loop the trigger 100×, parallelise, add stress, narrow timing windows, inject sleeps. A 50%-flake bug is debuggable; 1% is not — keep raising the rate until it's debuggable.
When you genuinely cannot build a loop
Stop and say so explicitly. List what you tried. Ask the user for: (a) access to whatever environment reproduces it, (b) a captured artifact (HAR file, log dump, core dump, screen recording with timestamps), or (c) permission to add temporary production instrumentation. Do not proceed to hypothesise without a loop.
Completion criterion — a tight loop that goes red
Phase 1 is done when the loop is tight and red-capable: you can name one command — a script path, a test invocation, a curl — that you have already run at least once (paste the invocation and its output), and that is:
Red-capable — it drives the actual bug code path and asserts the user's exact symptom, so it can go red on this bug and green once fixed. Not "runs without erroring" — it must be able to catch this specific bug.
Deterministic — same verdict every run (flaky bugs: a pinned, high reproduction rate, per above).
Fast — seconds, not minutes.
Agent-runnable — you can run it unattended; a human in the loop only via scripts/hitl-loop.template.sh.
If you catch yourself reading code to build a theory before this command exists, stop — jumping straight to a hypothesis is the exact failure this skill prevents. No red-capable command, no Phase 2.
Phase 2 — Reproduce + Minimise
Run the loop. Watch it go red — the bug appears.
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 high enough rate to debug against).
You have captured the exact symptom (error message, wrong output, slow timing) so later phases can verify the fix actually addresses it.
Minimise
Once it's red, shrink the repro to the smallest scenario that still goes red. Cut inputs, callers, config, data, and steps one at a time, re-running the loop after each cut — keep only what's load-bearing for the failure.
Why bother: a minimal repro shrinks the hypothesis space in Phase 3 (fewer moving parts left to suspect) and becomes the clean regression test in Phase 5.
Done when every remaining element is load-bearing — removing any one of them makes the loop go green.
Do not proceed until you have reproduced and minimised.
Phase 3 — Hypothesise
Generate 3–5 ranked hypotheses before testing any of them. Single-hypothesis generation anchors on the first plausible idea.
Each hypothesis must be falsifiable: state the prediction it makes.
Format: "If is the cause, then will make the bug disappear / will make it worse."
If you cannot state the prediction, the hypothesis is a vibe — discard or sharpen it.
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 saver. Don't block on it — 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:
Debugger / REPL inspection if the env supports it. One breakpoint beats ten logs.
Targeted logs at the boundaries that distinguish hypotheses.
Never "log everything and grep".
Tag every debug log with a unique prefix, e.g. [DEBUG-a4f2]. Cleanup at the end becomes a single grep. Untagged logs survive; tagged logs die.
Perf branch. For performance regressions, logs are usually wrong. Instead: establish a baseline measurement (timing harness, performance.now(), 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 is a correct seam for it.
A correct seam is one where the test exercises the real 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 can't replicate the chain that triggered the bug), a regression test there gives false confidence.
If no correct seam exists, that itself is the finding. Note it. The codebase architecture is preventing the bug from being locked down. Flag this for the next phase.
If a correct seam exists:
Turn the minimised repro into a failing test at that seam.
Watch it fail.
Apply the fix.
Watch it pass.
Re-run the Phase 1 feedback loop against the original (un-minimised) scenario.
Phase 6 — Cleanup + Post-mortem
Required before declaring done:
Original repro no longer reproduces (re-run the Phase 1 loop)
Regression test passes (or absence of seam is documented)
All [DEBUG-...] instrumentation removed — grep the prefix to confirm zero matches
Throwaway prototypes deleted (or moved to a clearly-marked debug location)
The hypothesis that turned out 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 good test seam, tangled callers, hidden coupling) hand off to the /improve-codebase-architecture skill with the specifics. Make the recommendation after the fix is in, not before — you have more information now than when you started.
Pitfalls
Jumping to a hypothesis before a red-capable loop exists. This is the #1 failure mode this skill prevents. If you catch yourself reading code to build a theory before you have a command that goes red, stop and go back to Phase 1.
Wrong bug. The loop reproduces a failure, but not the user's failure. Always confirm the symptom matches what the user described before proceeding.
Flaky loop. A 30-second loop that sometimes passes and sometimes fails is barely better than no loop. Tighten it: pin time, seed RNG, isolate filesystem, freeze network. Target seconds, not minutes.
Single-hypothesis anchoring. Generating one plausible idea and testing it immediately. Always generate 3–5 ranked, falsifiable hypotheses first.
Logging everything and grepping. This creates noise, not signal. Each probe must map to a specific prediction from Phase 3. Tag every debug log with a unique prefix like [DEBUG-a4f2] so cleanup is a single grep.
False-confidence regression test. Writing a regression test at a seam too shallow to replicate the real bug pattern. If no correct seam exists, document that — it's the finding.
Forgetting cleanup. Tagged debug logs and throwaway prototypes left in the codebase. Grep for [DEBUG- prefixes and delete or relocate prototypes before declaring done.
Skipping the user checkpoint in Phase 3. The ranked hypothesis list is a cheap checkpoint that can save significant time. Show it to the user before testing when possible.
Verification
Before declaring the bug fixed, verify each item:
Re-run the Phase 1 feedback loop — the original (un-minimised) scenario must now pass:
# Example: re-run the exact command from Phase 1
# Replace with your actual loop command
npm test -- --grep "bug-repro"
Expected: green / pass. If still red, the fix is incomplete.
Regression test passes (or absence of correct seam is documented):
# Run the regression test created in Phase 5
npm test -- --grep "regression-bug-<id>"
Expected: pass. If it fails, the fix doesn't cover the bug pattern.
No leftover instrumentation — grep for debug tags:
# PowerShell — search for any tagged debug logs
Select-String -Path "src\**\*.ts" -Pattern "\[DEBUG-"
Expected: zero matches. If matches found, remove them.
Throwaway prototypes removed — check for temporary harness files:
# List any throwaway debug scripts that should be deleted
Get-ChildItem -Path . -Filter "*debug*" -Recurse
Expected: only intentional, clearly-marked debug files remain.
Post-mortem recorded — the correct hypothesis is stated in the commit or PR message.
Related Skills
/improve-codebase-architecture — hand off here when the root cause involves architectural issues (no good test seam, tangled callers, hidden coupling). Make this recommendation after the fix is in.
