When to use

• Check for security vulnerabilities
• Performance optimization

Instructions

A comprehensive audit skill that covers two major domains:

1. Full Codebase Audit — UI, bugs, accessibility, performance, security, test coverage
2. System Prompt Audit — AI agent architecture review, prompt quality scoring, bottleneck analysis

Read the relevant reference file(s) based on what the user needs:

• Codebase audit → docs/references/codebase-audit.md
• System prompt audit → docs/references/prompt-audit.md
• Both → read both files

Step 1: Determine Audit Scope

Ask the user (or infer from context):

1. What is being audited? (repo URL/ZIP, pasted code, or system prompts)
2. What type of audit? (codebase, system prompts, or both)
3. Any specific flows, agents, or areas of concern?
4. Build commands, env vars, test credentials (for codebase audits)
5. Browsers/viewports to test (for UI audits)

If the user pastes code or prompts directly, begin immediately without asking.

Step 2: Run the Audit

Follow the instructions in the relevant reference file(s). Always:

• Produce a human-readable Executive Summary (top 5 findings + business impact)
• Produce a Technical Appendix (per-issue detail with file refs + line numbers)
• Produce a JSON summary (machine-readable, see schema in references/json-schema.md)
• Produce code patches or rewrites for all Critical/High issues
• Produce a 2-week sprint plan

Step 3: Output Format

Always structure output as:

# Audit Report: [Project/System Name]

## Executive Summary
[Top 5 findings, business impact, remediation roadmap]

## Technical Findings
[Per-issue detail — see references for full format]

## Patches & Fixes
[Copy-paste ready code diffs for Critical/High issues]

## CI Recommendations
[Automated checks to add]

## Sprint Plan
[2-week remediation roadmap]

## JSON Summary
[Machine-readable findings]

Severity Definitions

Reference Files

• docs/references/codebase-audit.md — Full instructions for repo/UI/bug/perf/security/a11y audit
• docs/references/prompt-audit.md — Full instructions for AI system prompt audit
• docs/references/json-schema.md — JSON output schema for both audit types
• docs/examples/codebase-report.md — Example output for a codebase audit
• docs/examples/prompt-report.md — Example output for a system prompt audit