dev-sect

Simulates a multi-persona discussion panel that debates and reviews code from several expert perspectives.

3|1|Updated Jan 12, 2026
One-click install
npx skills add https://github.com/nghyane/VSTEP --skill dev-sect-nghyane
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: dev-sect
Source: https://github.com/nghyane/VSTEP/tree/main/.agents/skills/dev-sect
Command: npx skills add https://github.com/nghyane/VSTEP --skill dev-sect-nghyane

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Getting a single AI response often means getting a single viewpoint, which can miss bugs, security holes, or design flaws. This Skill convenes multiple role-played expert personas who debate, challenge each other, and converge on a concrete decision before any code changes happen. ## Core Features & Use Cases - Multi-Persona Deliberation: Summons 3-5 expert personas (e.g., architecture, security, debugging) whose names self-describe their specialty, then runs one structured round of debate. - Adversarial Code Review: Personas are instructed to disagree bluntly rather than rubber-stamp, surfacing issues like race conditions, missing ownership checks, and absent transactions. - Decision Table Output: Ends with a concise decision table listing each task, its severity, and the agreed resolution, then stops and waits for the user's call. - Use Case: You suspect a grading submission endpoint has concurrency issues. Invoke the skill, and the personas identify the race condition, flag unverified file ownership, and agree on wrapping the logic in a database transaction. ## Quick Start Ask the team to review your code by saying something like: team đâu, review giúp đoạn code này.

Frequently Asked Questions about dev-sect

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get multiple AI perspectives on a code review?▼

Invoke the skill with a trigger phrase like 'team đâu' or 'review giúp' along with your code. It summons 3-5 expert personas such as an architecture specialist and a security specialist who debate the code and produce a joint decision table.

What kinds of issues can multi-persona code review catch?▼

The personas are instructed to challenge each other bluntly, which surfaces issues like race conditions, missing ownership verification, and absent database transactions. Each persona focuses on its own specialty such as architecture, security, or root-cause debugging.

Can the personas actually modify code or only discuss it?▼

Personas can implement changes, not just discuss. When the user says 'sửa', 'làm đi', or 'xuất thủ', the skill instructs immediate action rather than continued deliberation.

When should I not use a multi-persona discussion approach?▼

Avoid it for simple, single-answer tasks where one direct response is faster. The skill is designed for one debate round plus a decision, so it adds overhead that is only worthwhile for ambiguous or high-stakes technical decisions.

Does the discussion loop run indefinitely?▼

No. The skill enforces exactly one round of debate followed by a decision table, then stops and waits for the user to decide the next step. This prevents endless simulated discussion.