debugging-patterns

name: debugging-patterns description: "Use when a bug, flaky test, or runtime/build failure needs root-cause tracing and a nearby duplicate-pattern scan before any fix." allowed-tools: Read Grep Glob Bash LSP

Systematic Debugging

DIVERGENCE FROM superpowers:systematic-debugging: Forked. The four-phase root-cause discipline and the rationalization table are core debugging doctrine assumed here. CC10x ADDS: the feedback-loop-first gate (build a fast, deterministic, agent-runnable repro loop BEFORE any hypothesis), LSP-powered root-cause tracing, scenario playbooks, hypothesis confidence scoring, cognitive-bias and meta-debugging guidance, the Option-Zero (config-only fix) check, and the restart-investigation protocol. The bug-investigator agent owns the operational process (it enforces the feedback loop as a fail-closed gate); this skill is the advisory depth it loads.

Overview

Random fixes waste time and create new bugs. Quick patches mask underlying issues.

Core principle: ALWAYS find root cause before attempting fixes. Symptom fixes are failure.

This skill is advisory. It deepens investigation quality. It does not authorize local-only patches, guesswork, or "fix the line that crashed" thinking.

Reference Files

Read only the references needed for the current investigation:

• references/root-cause-playbooks.md for build/type failures, flaky tests, runtime crashes, browser errors, git bisect, and boundary tracing
• references/investigation-hygiene.md for context discipline, evidence logging, hypothesis tracking, restart protocol, and architectural escalation

The Iron Law

NO FIXES WITHOUT ROOT CAUSE INVESTIGATION FIRST

If you haven't completed Phase 1, you cannot propose fixes.

Feedback Loop FIRST (Before Any Hypothesis)

A hypothesis without a repro loop is a guess. Before you form H1 — before you read code to build a theory — build a fast, deterministic, agent-runnable pass/fail signal that turns red on this bug and that you can re-run on every iteration. The loop IS the evidence this skill already demands. Everything downstream (bisection, hypothesis testing, instrumentation) just consumes it; without it, no amount of staring at code will save you.

This is the advisory depth behind the bug-investigator agent's Feedback Loop Gate, which enforces it as a fail-closed gate: in the agent, no loop means no hypothesis (it returns BLOCKED, never advances to H1). The skill explains the technique; the agent makes it non-negotiable.

Spend disproportionate effort here. Be aggressive, be creative, refuse to give up.

Construction Ladder (try in rank order; stop at the first that is fast + deterministic)

1. Failing automated test (unit/integration/e2e) at whatever seam reaches the bug — best, because it is reusable as the RED regression test in Phase 4
2. curl/HTTP request with an asserted response (status/body diff) against a running dev server
3. CLI snapshot diff — run the command, diff stdout/stderr/exit code against a known-good snapshot
4. Headless browser script (Playwright/Puppeteer) — drives the real DOM/runtime, asserts on DOM/console/network (a real crash, not just types)
5. Trace replay — save a real request/payload/log/event to disk, replay it through the code path in isolation
6. Throwaway harness — a tiny script that calls the suspect function directly with a fixture input
7. Property/fuzz check — when the failing input is unknown, run many random inputs and watch for the failure mode
8. git bisect run — when the bug is a regression between two known states and a test exists
9. Differential old-vs-new — run the same input through the last-good revision beside HEAD (or two configs) and diff behavior
10. Human-in-the-loop — LAST resort: scripted manual steps the user runs and reports back, so the loop stays structured

Treat the Loop as a Product

Once you have a loop, do not settle for the first version. Sharpen it:

• Faster — sub-second beats sub-minute; you will run it dozens of times. Cache setup, skip unrelated init, narrow the scope.
• Sharper — assert the exact failing fact (the user's specific symptom), not a noisy superset or "didn't crash".
• More deterministic — same input → same red. Pin time, seed RNG, isolate filesystem, freeze network, remove jitter.

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

Flaky / Non-Deterministic Bugs

The goal is not a single clean repro — it is a higher reproduction RATE. Loop the trigger N times (for i in $(seq 1 N); do ...; done), record the hit rate (e.g. 3/50), and treat raising that rate as loop iteration: control the seed, force a schedule, add concurrency/load/stress, inject sleeps, narrow timing windows. A bug you can reproduce 3/50 times deterministically-on-replay beats one you cannot reproduce at all. A 50%-flake bug is debuggable; 1% is not — keep raising the rate until it is.

When You Genuinely Cannot Build a Loop

Stop and say so explicitly. Do not hypothesise into a vacuum. List each ladder rung you tried and why it failed, then ask the user for the one thing that would unblock the loop: (a) access to an environment that reproduces it, (b) a captured artifact (HAR file, log dump, core dump, full stack trace, failing input, screen recording with timestamps), or (c) permission to add temporary instrumentation in a live/prod path.

The loop comes FIRST. Only once it goes red do you proceed to the ranked, falsifiable hypotheses and confidence scoring below — those layer ON TOP of the loop, they do not replace it.

Quick Five-Step Process (Reference Pattern)

For rapid debugging, use this concise flow:

1. Capture error message and stack trace
2. Identify reproduction steps
3. Isolate the failure location
4. Implement minimal general fix
5. Verify solution works

Debugging techniques:

• Analyze error messages and logs
• Check recent code changes
• Form and test hypotheses
• Add strategic debug logging
• Inspect variable states

Root Cause Tracing Technique:

1. Observe symptom - Where does error manifest?
2. Find immediate cause - Which code produces the error?
3. Ask "What called this?" - Map call chain upward
4. Keep tracing up - Follow invalid data backward
5. Find original trigger - Where did problem actually start?

Never fix solely where errors appear—trace to the original trigger. After root cause is identified, scan for the same signature nearby before declaring success.

LSP-Powered Root Cause Tracing

Use LSP to trace execution flow systematically:

Systematic Call Chain Tracing:

1. localSearchCode("errorFunction") → get file + lineHint
2. lspGotoDefinition(lineHint=N) → see implementation
3. lspCallHierarchy(incoming, lineHint=N) → who calls this?
4. For each caller: lspCallHierarchy(incoming) → trace up
5. Continue until you find the root cause

CRITICAL: Always get lineHint from localSearchCode first. Never guess line numbers.

For each issue provide:

• Root cause explanation
• Evidence supporting diagnosis
• Specific code fix
• Testing approach
• Prevention recommendations

Scenario Playbooks

Read references/root-cause-playbooks.md when the failure matches one of these shapes:

• build or type breakage
• failing tests
• runtime crashes
• browser or console errors
• intermittent async bugs
• regressions where "it worked before"
• multi-component handoff failures

Keep this SKILL.md focused on the four-phase investigation workflow. Use the reference for concrete commands, boundary tracing patterns, and git-bisect recipes.

When to Use

Use for ANY technical issue:

• Test failures
• Bugs in production
• Unexpected behavior
• Performance problems
• Build failures
• Integration issues

Use this ESPECIALLY when:

• Under time pressure (emergencies make guessing tempting)
• "Just one quick fix" seems obvious
• You've already tried multiple fixes
• Previous fix didn't work
• You don't fully understand the issue

Don't skip when:

• Issue seems simple (simple bugs have root causes too)
• You're in a hurry (rushing guarantees rework)
• Manager wants it fixed NOW (systematic is faster than thrashing)

The Four Phases

You MUST complete each phase before proceeding to the next.

Phase 1: Root Cause Investigation

BEFORE attempting ANY fix:

1. Read Error Messages Carefully

   • Don't skip past errors or warnings
   • They often contain the exact solution
   • Read stack traces completely
   • Note line numbers, file paths, error codes

2. Reproduce Consistently — build the feedback loop FIRST

   • Can you trigger it reliably?
   • What are the exact steps?
   • Does it happen every time?
   • If not reproducible → gather more data, don't guess
   • This is where you build the agent-runnable repro loop (see Feedback Loop FIRST above). Use the construction ladder; do NOT proceed to Phase 3 hypotheses until you have a red-capable loop. For flaky bugs, chase a higher reproduction rate, not a single clean hit.

3. Check Recent Changes

   • What changed that could cause this?
   • Git diff, recent commits
   • New dependencies, config changes
   • Environmental differences

4. Gather Evidence in Multi-Component Systems

   WHEN system has multiple components (CI → build → signing, API → service → database):

   BEFORE proposing fixes, add diagnostic instrumentation:

   For EACH component boundary:
     - Log what data enters component
     - Log what data exits component
     - Verify environment/config propagation
     - Check state at each layer
   
   Run once to gather evidence showing WHERE it breaks
   THEN analyze evidence to identify failing component
   THEN investigate that specific component
   

   For a concrete boundary-tracing recipe, read references/root-cause-playbooks.md.

5. Trace Data Flow

   WHEN error is deep in call stack:

   • Where does bad value originate?
   • What called this with bad value?
   • Keep tracing up until you find the source
   • Fix at source, not at symptom

6. Trace configuration to its consumer (third-party SDKs, env, import time)

   BEFORE proposing replacements, migrations, or large refactors:

   • Option Zero: Can this be fixed with only a configuration or environment change? State and test that hypothesis before you propose code rewrites. Many integration bugs are wiring problems, not architecture problems.
   • When an env var is validated in application config but never passed into library constructors, ask who reads it? Third-party SDKs often read process.env (or language equivalents) at module import / load time—not through your app's DI or config objects. The fix may be setting or correcting env, not changing application code.
   • Before proposing to replace a third-party SDK, spend a short cycle on how it is configured: package README, official docs, or the installed source under node_modules (or vendor path). The failure may be a wrong endpoint or flag, not a need for a different library.
   • When code comments describe import-time behavior or env-var dependencies, treat them as investigation breadcrumbs, not decoration—follow them to the real consumer.

Phase 2: Pattern Analysis

Find the pattern before fixing:

1. Find Working Examples

   • Locate similar working code in same codebase
   • What works that's similar to what's broken?

2. Compare Against References

   • If implementing pattern, read reference implementation COMPLETELY
   • Don't skim - read every line
   • Understand the pattern fully before applying

3. Identify Differences

   • What's different between working and broken?
   • List every difference, however small
   • Don't assume "that can't matter"

4. Understand Dependencies

   • What other components does this need?
   • What settings, config, environment?
   • What assumptions does it make?

Phase 3: Hypothesis and Testing

Scientific method:

1. Form Single Hypothesis

   • State clearly: "I think X is the root cause because Y"
   • Write it down
   • Be specific, not vague

2. Test Minimally

   • Make the SMALLEST possible change to test hypothesis
   • One variable at a time
   • Don't fix multiple things at once

3. Verify Before Continuing

   • Did it work? Yes → Phase 4
   • Didn't work? Form NEW hypothesis
   • DON'T add more fixes on top

4. When You Don't Know

   • Say "I don't understand X"
   • Don't pretend to know
   • Ask for help
   • Research more

Hypothesis Quality Criteria

Falsifiability Requirement: A good hypothesis can be proven wrong. If you can't design an experiment to disprove it, it's not useful.

Bad (unfalsifiable):

• "Something is wrong with the state"
• "The timing is off"
• "There's a race condition somewhere"

Good (falsifiable):

• "User state resets because component remounts when route changes"
• "API call completes after unmount, causing state update on unmounted component"
• "Two async operations modify same array without locking, causing data loss"

The difference: Specificity. Good hypotheses make specific, testable claims.

Hypothesis Confidence Scoring

Track multiple hypotheses with confidence levels:

H1: [hypothesis] — Confidence: [0-100]
    Evidence for: [what supports this]
    Evidence against: [what contradicts this]
    Next test: [what would raise or lower confidence]

H2: [hypothesis] — Confidence: [0-100]
    Evidence for: [...]
    Evidence against: [...]
    Next test: [...]

H3: [hypothesis] — Confidence: [0-100]
    Evidence for: [...]
    Evidence against: [...]
    Next test: [...]

Scoring guidance:

Rules:

• Always maintain 2-3 hypotheses until one reaches 80+
• Update confidence after EVERY piece of new evidence
• Never proceed to fix with highest hypothesis below 50

Cognitive Biases in Debugging

Meta-Debugging: Your Own Code

When debugging code you wrote, you're fighting your own mental model.

Why this is harder:

• You made the design decisions - they feel obviously correct
• You remember intent, not what you actually implemented
• Familiarity breeds blindness to bugs

The discipline:

1. Treat your code as foreign - Read it as if someone else wrote it
2. Question your design decisions - Your implementation choices are hypotheses, not facts
3. Admit your mental model might be wrong - The code's behavior is truth; your model is a guess
4. Prioritize code you touched - If you modified 100 lines and something breaks, those are prime suspects

The hardest admission: "I implemented this wrong." Not "requirements were unclear" - YOU made an error.

When to Restart Investigation

Consider starting over when:

1. 2+ hours with no progress - You're likely tunnel-visioned
2. 3+ "fixes" that didn't work - Your mental model is wrong
3. You can't explain the current behavior - Don't add changes on top of confusion
4. You're debugging the debugger - Something fundamental is wrong
5. The fix works but you don't know why - This isn't fixed, this is luck

Restart protocol:

1. Close all files and terminals
2. Write down what you know for certain
3. Write down what you've ruled out
4. List new hypotheses (different from before)
5. Begin again from Phase 1

Phase 4: Implementation

Fix the root cause, not the symptom:

1. Create Failing Test Case

   • Simplest possible reproduction
   • Automated test if possible
   • One-off test script if no framework
   • MUST have before fixing

2. Implement Single Fix

   • Address the root cause identified
   • ONE change at a time
   • No "while I'm here" improvements
   • No bundled refactoring

3. Verify Fix

   • Test passes now?
   • No other tests broken?
   • Issue actually resolved?

4. If Fix Doesn't Work

   • STOP
   • Count: How many fixes have you tried?
   • If < 3: Return to Phase 1, re-analyze with new information
   • If >= 3: STOP and question the architecture (step 5 below)
   • DON'T attempt Fix #4 without architectural discussion

5. If 3+ Fixes Failed: Question Architecture

   Pattern indicating architectural problem:

   • Each fix reveals new shared state/coupling/problem in different place
   • Fixes require "massive refactoring" to implement
   • Each fix creates new symptoms elsewhere

   STOP and question fundamentals:

   • Is this pattern fundamentally sound?
   • Are we "sticking with it through sheer inertia"?
   • Should we refactor architecture vs. continue fixing symptoms?

   Discuss with the user before attempting more fixes

   This is NOT a failed hypothesis - this is a wrong architecture.

Red Flags - STOP and Follow Process

If you catch yourself thinking:

• "Quick fix for now, investigate later"
• "Just try changing X and see if it works"
• "Add multiple changes, run tests"
• "Skip the test, I'll manually verify"
• "It's probably X, let me fix that"
• "I don't fully understand but this might work"
• "Pattern says X but I'll adapt it differently"
• "Here are the main problems: [lists fixes without investigation]"
• Proposing solutions before tracing data flow
• Proposing SDK replacement or a large migration before ruling out a config-only fix (Option Zero) and tracing who reads env / config at import time
• "One more fix attempt" (when already tried 2+)
• Each fix reveals new problem in different place

ALL of these mean: STOP. Return to Phase 1.

If 3+ fixes failed: Question the architecture (see Phase 4.5)

User's Signals & Rationalizations

The user-redirection signals ("Stop guessing", "Ultrathink this", "We're stuck?") and the excuse/reality rationalization table are core systematic-debugging discipline — assumed, not repeated here (see superpowers:systematic-debugging). The operational rule: any of these means STOP and return to Phase 1; simple/emergency bugs still have root causes and process is faster than guess-and-check; 3+ failed fixes means an architectural problem, not "one more attempt".

Quick Reference

When Process Reveals "No Root Cause"

If systematic investigation reveals issue is truly environmental, timing-dependent, or external:

1. You've completed the process
2. Document what you investigated
3. Implement appropriate handling (retry, timeout, error message)
4. Add monitoring/logging for future investigation

But: 95% of "no root cause" cases are incomplete investigation.

Output Format

## Bug Investigation

### Phase 1: Evidence Gathered
- **Error**: [exact error message]
- **Stack trace**: [relevant lines]
- **Reproduction**: [steps to reproduce]
- **Recent changes**: [commits/changes]

### Phase 2: Pattern Analysis
- **Working example**: [similar working code]
- **Key differences**: [what's different]

### Phase 3: Hypothesis
- **Theory**: [I think X because Y]
- **Test**: [minimal change made]
- **Result**: [confirmed/refuted]

### Phase 4: Fix
- **Root cause**: [actual cause with evidence]
- **Change**: [summary of fix]
- **File**: [path:line]
- **Regression test**: [test added]

### Verification
- Test command: [command] → exit 0
- All tests: PASS
- Functionality: Restored