brutal-exhaustive-audit

Brutal Exhaustive Audit

Overview

This is the "be exhaustive, be brutal, check everything" skill codified into an enforceable process. AI's context window makes it forget things, skip files, and declare victory early. This skill prevents all of that through mandatory multi-pass verification with explicit checklists.

Core principle: Every file checked. Every route visited. Every flow walked. No exceptions, no shortcuts, no "it's probably fine."

The Iron Law

NO FILE UNCHECKED. NO ROUTE UNVISITED. NO FLOW UNTESTED. NO CLAIM WITHOUT EVIDENCE.
IF YOU HAVEN'T EXPLICITLY VERIFIED IT, IT IS BROKEN UNTIL PROVEN OTHERWISE.

When to Use

• When you say "be exhaustive" or "be brutal" or "check everything"
• When you don't trust the AI did a complete job
• When previous fixes introduced new problems
• After a large-scale refactoring or migration
• After API integration to verify nothing was missed
• Before a release or demo
• When the product "should work" but something feels wrong

When NOT to Use

• For a quick single-file fix (use fix-issue)
• For understanding a new codebase (use codebase-mapping)

The Five Passes

THIS AUDIT HAS FIVE MANDATORY PASSES.
EACH PASS PRODUCES A CHECKLIST.
EVERY ITEM ON EVERY CHECKLIST MUST BE EXPLICITLY VERIFIED.
SKIPPING A PASS IS NOT AN OPTION.

Pass 1: Build & Static Analysis (THE FOUNDATION)

IF IT DOESN'T BUILD, NOTHING ELSE MATTERS.

1. CLEAN BUILD:
   □ Delete all build artifacts (dist, .next, build, __pycache__)
   □ Delete node_modules and reinstall (npm ci or pip install -r requirements.txt)
   □ Run the build from scratch
   □ Record: Exit code, warnings, errors
   □ EVERY warning matters — list them all

2. TYPE CHECK (if applicable):
   □ Run TypeScript compiler with --noEmit (or mypy/pyright)
   □ Record: EVERY type error, not just the count
   □ Categorize: Critical (broken logic) vs Minor (missing annotation)

3. LINT CHECK:
   □ Run the full linter
   □ Record: EVERY lint error
   □ Flag: Any rules that are disabled or overridden (eslint-disable, noqa, etc.)

4. DEPENDENCY CHECK:
   □ Are there unused dependencies?
   □ Are there missing dependencies?
   □ Are there security vulnerabilities? (npm audit, pip-audit)
   □ Are there peer dependency conflicts?

5. ENVIRONMENT CHECK:
   □ Are all required environment variables documented?
   □ Are all required environment variables set (or have defaults)?
   □ Are there environment-specific configurations that could break?

PASS 1 RESULT:
| Check | Status | Issues |
|-------|--------|--------|
| Clean Build | ✅/❌ | [details] |
| Type Check | ✅/❌ | [N errors listed] |
| Lint | ✅/❌ | [N errors listed] |
| Dependencies | ✅/❌ | [details] |
| Environment | ✅/❌ | [details] |