agentleFS
joeloverbeckjoeloverbeck/skill-evidence/.claude/skills/diagnosing-bugs/SKILL.md

diagnosing-bugs

Skill0 starsChanged 3 months agoMITView on GitHub

Diagnose hard bugs and performance regressions, and repair them when authorized. Use when the user asks to diagnose or debug, reports something broken/throwing/failing/slow, or explicitly asks to fix a suspected bug; default to diagnosis-only unless repair is requested.

193 other projects use a version of this, 160 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use setup-ts-deep-modules, loop-me and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • liufeng1976Skill

    setup-ts-deep-modules

    Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.

    Also in 43 projects that use this

  • RoentekSkill

    loop-me

    Grill me about specs for the workflows I want to build, within this workspace.

    Also in 55 projects that use this

  • LesaniSkill

    writing-beats

    Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.

    Also in 58 projects that use this

  • AWSSkill

    to-questionnaire

    Turn a decision you can't fully answer into a questionnaire for someone else to fill in.

    Also in 88 projects that use this

  • LesaniSkill

    writing-shape

    Take a markdown file of raw material and shape it into an article through a conversational session — drafting candidate openings, growing the piece paragraph by paragraph, arguing about format (lists, tables, callouts, quotes) at each step. Use when the user has a pile of notes, fragments, or a rough draft and wants help turning it into something publishable.

    Also in 56 projects that use this

  • AWSSkill

    wizard

    Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.

    Also in 122 projects that use this

  • +1 more from Lesani
  • +1 more from aws

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. 12 are common, 3 are distinctive, and 1 is very specific to this project.

  • 12

    Common

    Most of the 182 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    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 …

    line 10
  2. Common:

    If the redacted output is not enough to diagnose the bug, say so and ask the user.

    line 16
  3. Distinctive:

    When scope is ambiguous, default to diagnosis only.

    line 25
  4. Distinctive:

    Property / fuzz loop: If the bug is "sometimes wrong output", run a budgeted random sample, record the observed failure rate, and retain …

    line 41
  5. Common:

    Bisection harness: If the bug appeared between two known states (commit, dataset, version), automate "boot at state X, check, repeat" so …

    line 42
  6. Common:

    Differential loop: Run the same input through old-version vs new-version (or two configs) and diff outputs.

    line 43
  7. Common:

    Treat the loop as a product.

    line 50
  8. Common:

    If you catch yourself reading code to build a theory before this command exists, stop — jumping straight to a hypothesis is the exact …

    line 75
  9. Common:

    Run the loop. Watch it go red — the bug appears.

    line 79
  10. Distinctive:

    Before calling the red signal "the bug", record one row per distinct failure:

    line 89
  11. Common:

    Do not proceed until you have reproduced and minimised.

    line 111
  12. Common:

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

    line 123
  13. Common:

    Debugger / REPL inspection: if the env supports it. One breakpoint beats ten logs.

    line 133
  14. Common:

    Never "log everything and grep".

    line 135
  15. Common:

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

    line 151
  16. Common:

    If a correct seam exists:

    line 157

“Common” is measured: kept by at least 6 in 10 of the 182 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Diagnosing Bugs

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.

Redact

This skill has you show commands, outputs and captured artifacts. Redact every secret first — write <REDACTED> in its place. Build loops against env vars, so the credential stays in the environment rather than in what you show. Captured artifacts carry auth headers: quote only the lines that carry the signal.

If the redacted output is not enough to diagnose the bug, say so and ask the user.

Request Mode

Classify the request before any mutation:

  • Diagnosis only — run Phases 1–4 as needed, then stop with the failure-surface classification, supported causal finding, confidence and remaining unknowns, and repair options. Do not enter Phase 5.
  • Diagnosis and repair — run the full loop only when the user explicitly asked to fix or change the system, or later authorizes one of the repair options.

When scope is ambiguous, default to diagnosis only. Permission for temporary instrumentation authorizes only that instrumentation, not the eventual repair.

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; 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.

Ways to construct one — try them in roughly 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 network request / payload / event log to disk; replay it through the code path in isolation.
  6. 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.
  7. Property / fuzz loop. If the bug is "sometimes wrong output", run a budgeted random sample, record the observed failure rate, and retain failing seeds as fixtures.
  8. Bisection harness. If the bug appeared between two known states (commit, dataset, version), automate "boot at state X, check, repeat" so you can git bisect run it.
  9. Differential loop. Run the same input through old-version vs new-version (or two configs) and diff outputs.
  10. 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.

A tight feedback loop materially reduces uncertainty; it does not establish the cause or prove the repair.

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 loop is tight when it is specific and deterministic enough to compare probes, and cheap enough to rerun after every meaningful change within the task's diagnostic budget. Record its wall time and failure rate instead of applying a universal duration cutoff.

Non-deterministic bugs

The goal is not a clean repro but enough observed failures to compare hypotheses within the available time or cost budget. Measure the baseline rate, estimate expected failures per batch (attempts × rate), and choose a batch that should yield several observations. A low raw rate remains usable when attempts are cheap and agent-runnable; when expected observations are too sparse, parallelise, add stress, narrow timing windows, or inject sleeps until the loop becomes informative. Record the rate, batch size, run cost, and observed failures instead of applying a universal percentage cutoff.

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 redacted 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 (show the invocation and its output, redacted), 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).
  • Affordable — cheap enough to rerun after every meaningful probe within the task's diagnostic budget; record its wall time and per-batch cost.
  • 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.

Classify the failure surface

Before calling the red signal "the bug", record one row per distinct failure:

FieldRequired content
Observed signalExact error, wrong output, timing, or rejection
Controlling contractAssertion, schema, protocol, specification, or user expectation that makes it red
ExpectedRequired value or behavior
ActualObserved value or behavior
Fault classimplementation defect, input defect, protocol prerequisite, substantive failure, or unknown
Causal impactWhat this failure blocks or invalidates
Unaffected scopeBehavior, evidence, or verdicts this failure does not decide

A validator rejection may be intended enforcement rather than a validator defect. Keep separate rows when one run exposes failures in different layers. Show the classification and a plain-language causal chain to the user before recommending repair; if the class remains unknown, do not imply that an implementation is defective.

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

Unless the direct-proof exception below applies, generate 3–5 ranked hypotheses before testing any of them. Single-hypothesis generation anchors on the first plausible idea.

Direct-proof exception. Skip the ranked set only when the red signal names the exact mismatch, the controlling assertion is current, and direct source bytes establish one complete causal chain. Record red signal → controlling assertion → expected → actual → mechanism plus one observation that would falsify the chain, and show it to the user. If any link is inferred, ambiguous, or contested, use the ranked-hypothesis path.

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 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:

  1. Debugger / REPL inspection if the env supports it. One breakpoint beats ten logs.
  2. Targeted logs at the boundaries that distinguish 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 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 — Authorized fix + regression test

Enter this phase only on a diagnosis-and-repair request. A diagnosis-only run ends after the supported causal finding and repair options are delivered.

Artifact lifecycle gate

Before editing, read the repository or project convention that owns the target artifact and classify it as a mutable working surface or frozen, published, or custodied evidence. Mutate only a mutable owner. For frozen evidence, preserve the producing bytes, hashes, invalid attempts, and superseded artifacts, then use the owner's versioning or supersession route and rerun every downstream check it requires.

A stronger governing or companion workflow's custody, freeze, or supersession protocol takes precedence and is followed unchanged.

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:

  1. Turn the minimised repro into a failing test at that seam.
  2. Watch it fail.
  3. Apply the authorized fix at the earliest mutable owner, or create the governed superseding or versioned artifact when the lifecycle owner forbids in-place change.
  4. Watch it pass.
  5. Re-run the Phase 1 feedback loop against the original (un-minimised) scenario.

Phase 6 — Authorized-repair cleanup + post-mortem

Required before declaring an authorized repair 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)
  • Truly throwaway prototypes deleted (or moved to a clearly-marked debug location); captured traces, invalid attempts, frozen artifacts, and superseded evidence retained whenever the governing custody or provenance contract requires them
  • The proven causal chain or 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) tell the user to run 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.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.