Menu
Diagnose bugs, test failures, build failures, unexpected behavior, and performance regressions before proposing fixes. Build a tight reproduction loop, find the root cause, fix it once, and lock it down with a regression test.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when specs are approved (phase 4 approved in docs/spp/pipeline-state.md) OR when the user directly asks to turn a spec into an implementation plan outside a running pipeline (e.g. "turn this spec into an implementation plan") - writes implementation plans as bite-sized tasks to docs/spp/05-plans/ for subagent-driven execution
Use after completing a task or major feature, before merging, or when the user asks to review a branch, PR, or work-in-progress diff. Run independent Standards and Spec reviews so clean code cannot hide the wrong behavior and correct behavior cannot hide broken standards.
Use when implementing any feature or bugfix, before writing implementation code
Use when starting any conversation - routes each request to a single SPP skill by default, running the full idea-to-deploy pipeline only when the user explicitly asks for "pipeline"
Use when a product is ready to ship, OR when the user directly asks how or where to deploy or to write a deploy runbook - chooses a deploy strategy with the owner in cost-and-consequence terms, then executes it into a repeatable runbook. Self-checks on disk for a fixed release version first โ it warns and offers to fix it, it does not block.
Use when the user describes a product idea, OR directly asks to capture or write up a product idea brief - captures an idea brief through a one-question-at-a-time interview. Runnable standalone; if docs/spp/pipeline-state.md is absent it starts a fresh journal, if present it appends.
| name | systematic-debugging |
| description | Diagnose bugs, test failures, build failures, unexpected behavior, and performance regressions before proposing fixes. Build a tight reproduction loop, find the root cause, fix it once, and lock it down with a regression test. |
Adapted from mattpocock/skills, MIT. SPP additions: proportional execution, component-boundary evidence, recent-change analysis, and architecture escalation after repeated failed fixes.
A discipline for hard bugs. Skip phases only when explicitly justified.
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.
Apply the process proportionally, but never skip the signal. For an obvious syntax, configuration, or build error, the existing failing command may already be the tight loop; capture its exact failure and continue. For a hard or intermittent bug, invest heavily in making the loop deterministic and fast.
This is the skill. Everything else is mechanical. If you have a tight pass/fail signal for the bug โ one that goes red on this bug โ you will find the cause; bisection, hypothesis-testing, and instrumentation all just consume it. 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.
git bisect run it.scripts/hitl-loop.template.sh so the loop is still structured. Captured output feeds back to you.Build the right feedback loop, and the bug is 90% fixed.
When the path crosses components (CI โ build โ deploy, client โ API โ service โ database), capture evidence at every boundary before choosing a suspect:
Run the path once with targeted instrumentation. Use the evidence to identify the failing boundary, then minimise inside that component. Do not patch the first layer that happens to surface the error.
Treat the loop as a product. Once you have a loop, tighten it:
A 30-second flaky loop is barely better than no loop; a 2-second deterministic one is tight โ a debugging superpower.
The goal is not a clean repro but a higher reproduction rate. Loop the trigger 100ร, parallelise, add stress, narrow timing windows, inject sleeps. A 50%-flake bug is debuggable; 1% is not โ keep raising the rate until it's debuggable.
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.
Phase 1 is done when the loop is tight and red-capable: you can name one command โ a script path, a test invocation, a curl โ that you have already run at least once (paste the invocation and its output), and that is:
scripts/hitl-loop.template.sh.If you catch yourself reading code to build a theory before this command exists, stop โ jumping straight to a hypothesis is the exact failure this skill prevents. No red-capable command, no Phase 2.
Run the loop. Watch it go red โ the bug appears.
Confirm:
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 โ 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 โ removing any one of them makes the loop go green.
Do not proceed until you have reproduced and minimised.
Before ranking hypotheses, inspect recent changes and find the closest working example in the same codebase. Compare the working and failing paths explicitly; small configuration, dependency, data-shape, and environment differences count as evidence.
Generate 3โ5 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 โ 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 โ proceed with your ranking if the user is AFK.
Each probe must map to a specific prediction from Phase 3. Change one variable at a time.
Tool preference:
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.
Write the regression test before the fix โ 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 and a blocker. The codebase architecture is preventing the bug from being locked down. Stop before the fix and agree on a real seam or the architectural change that creates one; do not mark the bug fixed without a regression test that catches the original pattern.
If a correct seam exists:
Required before declaring done:
[DEBUG-...] instrumentation removed (grep the prefix)Then ask: what would have prevented this bug? If the answer involves architectural change (no good test seam, tangled callers, hidden coupling), use the codebase-design skill to frame the seam and module changes. Make the recommendation after the fix is in, not before โ you have more information now than when you started.
If 3 attempted fixes have failed, stop before attempt 4. Re-evaluate the architecture with the user: repeated fixes exposing new coupling or shared state are evidence that the current design, not the latest symptom, may be wrong.