Named-Persona Adversarial Review — Development

Named-Persona Adversarial Review performs code review through the documented philosophies of real engineers (e.g., Torvalds, Thompson, Carmack, Jobs, Cagan), grounding each finding in a sourced principle with a confidence level. It produces a structured review report with severity levels, concurrence promotion, and an integrity check to avoid hallucinated attributions.

engineering claude-code

📂 View skill on GitHub ↗ 🚀 How to use

✨ What it does

Grounds each review in real, documented principles from references/persona_principles.md.
Enforces attribution discipline: cites principles with confidence levels, never fabricates quotes.
Mandatory product persona each round to catch UX issues.
Severity levels (BLOCKER, CRITICAL, WARNING, NOTE) with promotion on concurrence.
Zero-finding burden: requires naming principles satisfied if no issues found.
Integrity check (Feynman) to prevent self-deception and generic advice dressed in famous names.

🎯 When to use it

When automated review findings feel generic and you want sourced, specific critiques.
For PRs with architectural, UX, or security impact that need deeper review.
Before submitting a self-authored PR for pre-submit hardening.
When reviewing methodologies or documentation where product personas add value.

🚀 How to use

Trigger with phrases like 'review this PR with real engineers' or 'named persona review'. Provide the code diff or PR context. The skill selects personas based on routing rules (e.g., code correctness → Torvalds + Carmack + Jobs). Example prompts:

Review this PR with named personas: Torvalds, Thompson, and Jobs.
Named persona review on this auth refactor.

📄 Output: A structured review report with findings per persona, severity levels, and a verdict, posted as a PR comment or saved to a file.

📦 Add this skill to Claude Code

# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills

# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/engineering-team/skills/named-persona-adversarial-review .claude/skills/named-persona-adversarial-review

Skill source: engineering-team/skills/named-persona-adversarial-review/SKILL.md

⚠️ Good to know

Requires grounding in references/persona_principles.md or web access; without it, personas cannot be used. Not for low-impact or throwaway code.

❓ FAQ

What if I can't ground a persona's philosophy?

Drop that persona; never fabricate a citation. The skill requires grounding before role-play.

How many personas are used per review?

Three: two engineers and one product persona (mandatory).

What is the zero-finding burden?

If no issues are found, you must name at least three principles the code satisfies and explain how.

🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.