Audit Fix

---
description: Audit code → fix all findings → verify → repeat until clean
argument-hint: "[scope: empty=uncommitted, commit -N, path]"
---

# Audit-Fix Loop

Audit code, fix all findings, verify fixes. Loop until zero findings remain.

## Input

```text
$ARGUMENTS

Phase 1: Scope

Parse $ARGUMENTS to determine what to audit:

Input	Scope
(empty)	Uncommitted changes (git diff HEAD --name-only)
staged	Staged changes (git diff --cached --name-only)
commit -1	Last commit (git diff HEAD~1 --name-only)
commit -N	Last N commits (git diff HEAD~N --name-only)
path/to/file	Specific file or directory

If no changed files found: report "Nothing to audit" and STOP.

Phase 2: Audit

Read each changed file and analyze across these dimensions:

1. Correctness & logic — is the code logically sound? No patching around symptoms.
2. Edge cases — boundary conditions, null/empty, Unicode/CJK, concurrent access
3. Security — no vulnerabilities (injection, XSS, path traversal, SQL injection)
4. Duplicate code — copy-paste patterns, repeated logic that should be unified
5. Dead code — unused imports, unreachable branches, orphaned functions
6. Shortcuts & patches — workarounds, TODO markers, band-aids, bypass flags
7. Project compliance — adherence to rules in .claude/rules/*.md and AGENTS.md

Report EVERY issue as: file:line | severity (Critical/High/Medium/Low) | issue | fix

Phase 3: Fix All

Fix every finding — Critical, High, Medium, and Low. No exceptions, no deferrals.

Rules:

• Fix the root cause, not the symptom.
• If fixing introduces new code, apply the same audit dimensions mentally before moving on.
• Keep diffs minimal and focused.
• Follow project conventions (.claude/rules/*.md).

Phase 4: Verify

Re-read each fixed file:

1. Confirm the original issue is resolved
2. Check for regressions or new issues introduced by the fix
3. Report any remaining findings

Phase 5: Loop or Exit

• Zero findings (all severities): audit passes. Report "Clean" to user.
• Findings remain and iteration < 3: go back to Phase 3, fix, then Phase 4 verify again.
• 3 iterations exhausted: STOP. List all remaining findings. Do not declare clean.

Phase 6: Gate (optional)

If the loop exited clean, offer to run the project's full test/gate command. Report pass/fail to user. ```