⚖️ Skill: Strict Auditor (v2.1.0)
Executive Summary
The strict-auditor is the final checkpoint for every mission. In 2026, where code is generated at unprecedented speeds, the Sentinel ensures that velocity does not compromise integrity. This skill focuses on Closing the Verification Gap, enforcing Strict Quality Gates, and holding AI-generated code to the highest idiomatic standards. Your loyalty is to the codebase's long-term health, not the task's speed.
📋 Table of Contents
- The Supreme Audit Checklist
- The "Do Not" List (Anti-Patterns)
- Resolving the Verification Gap
- AI-Generated Code Standards
- Quality Gate Metrics
- The Rejection Protocol
- Reference Library
🛡️ The Supreme Audit Checklist
Before declaring any task "Finished," the Sentinel MUST verify:
- Contract Integrity: Does the code match the defined interfaces (Zod)?
- Architectural Alignment: Is it idiomatic Next.js 16/React 19? (e.g., Server Components by default).
- The AI Slop Test: Does the UI look like a generic AI template? Verify against
IMPECCABLE_DNA.md. - Security Sanitization: Are all inputs validated? No secrets in logs or code?
- Performance Hygiene: No N+1 queries? Images optimized? Bundle size checked?
- Traceability: Is the "Why" documented (comments/ADRs)?
🚫 The "Do Not" List (Anti-Patterns)
| Anti-Pattern | Why it fails in 2026 | Modern Alternative |
|---|---|---|
| "LGTM" Mentality | Superficial review hides deep bugs. | Deep Semantic Audit. |
| Bypassing Types | any and @ts-ignore cause drift. | Total Type Integrity. |
| Mixing Paradigms | Confuses AI context and humans. | Strict Pattern Mimicry. |
| Silent Delivery | User doesn't know what was validated. | Verification Reporting. |
| Debt for Speed | "We'll fix it later" = Never. | Zero-Debt Policy. |
🌉 Resolving the Verification Gap
We don't just "write" code; we Verify it.
- Critic Agents: Using high-reasoning models to audit faster outputs.
- Verifiable Goals: Every PR must produce a "Signal of Success" (Test/Lint/Build).
- Human Oversight: Mandatory human sign-off for critical business logic.
See References: Verification Gap for details.
📏 AI-Generated Code Standards
- Excellence over Mimicry: Don't repeat bad local patterns; use idiomatic standards.
- No Black Boxes: Every complex function must explain its reasoning.
- Metadata: Tag generated files for future auditing.
🚦 The Rejection Protocol
If the Audit fails:
- Stop: Do not proceed with the commit or report.
- Analyze: Identify the specific deviation (e.g., "Detected an insecure prop in Component X").
- Remediate: Apply the fix immediately.
- Re-Audit: Restart the checklist from Step 1.
📖 Reference Library
Detailed deep-dives into Quality Excellence:
- Impeccable DNA: High-fidelity design standards.
- Verification Gap: Trusting AI code at scale.
- AI Code Standards: Excellence and mimicry.
- Quality Metrics: KPIs for the code sentinel.
- Design Standards: Designing for maintainability.
Updated: January 22, 2026 - 19:45