Executive Reference & Commands

Skill Brain turns general-purpose AI coding agents into disciplined senior engineering partners. Rather than executing monolithic, hallucinated passes, the agent decomposes requests across 6 categories, confirms its plan, loads at most 2–3 precision skills simultaneously, and verifies all work with raw CLI evidence.

Dollar Commands & Shorthand Triggers ($) & Planners

You can invoke any category, dedicated pre-implementation planner, or specialized capability instantly using dollar commands ($):

Dynamic Responsive Sizing & Universal Device Support

A critical mandate of Skill Brain is that every UI component and page must adapt dynamically to every screen size — from compact 320px mobile phones to 2560px ultra-wide desktop monitors. We divide this into two specialized layout disciplines, unified by a mathematical clamp engine.

The Mathematical Fluid Clamp Engine (`dynamic-ui-engine`)

Hardcoded pixel widths and arbitrary media query jumps are replaced with continuous fluid clamp mathematics that calculate optimal values across any screen width:

/* Universal Dynamic Responsive Tokens */
:root {
  /* Fluid Spacing Scale: 320px -> 1440px */
  --space-xs:  clamp(0.75rem, 0.7rem + 0.25vw, 0.875rem);
  --space-sm:  clamp(1rem, 0.9rem + 0.5vw, 1.25rem);
  --space-md:  clamp(1.5rem, 1.3rem + 1vw, 2rem);
  --space-lg:  clamp(2rem, 1.7rem + 1.5vw, 3rem);

  /* Fluid Typography Scale */
  --font-size-base: clamp(0.9375rem, 0.9rem + 0.19vw, 1.0625rem);
  --font-size-lg:   clamp(1.125rem, 1.05rem + 0.38vw, 1.375rem);
  --font-size-hero: clamp(2.25rem, 1.8rem + 2.25vw, 4rem);
}

Backend Responsive UI Contracts (`responsive-contract-enforcer`)

Responsive design is not merely a CSS concern — it begins at the backend contract layer. Backend engineers and API designers must honor these three contract rules:

1. No Hardcoded Dimensions in Payloads: APIs must never return fixed pixel coordinates (e.g. width: 1200). Send semantic layout tokens (layout: "single-stack" | "auto-fill").
2. Aspect Ratio & SrcSet Metadata: Every media entity returned by an API must include explicit aspect ratios (e.g. "16/9") and multi-resolution variant arrays to eliminate Cumulative Layout Shift (CLS).
3. Adaptive Pagination: Cursor-based stream pagination for mobile feeds; indexed pagination with total counts for desktop data tables.

The Virtual Memory Engine

Context windows are volatile and ephemeral; files on disk are permanent and auditable. The Virtual Memory Engine ensures complex software builds never drift or lose context across long sessions.

.agents/memory/
├── plans/
│   └── main-plan.md       ← Active master plan
├── phases/
│   ├── phase-01-scaffold.md
│   ├── phase-02-core-logic.md
│   └── phase-03-ui-polish.md
├── archive/               ← Completed & verified
│   └── 2026-11-02-auth/
└── decisions/             ← Permanent ADRs
    └── ADR-001-jwt.md

The 4-Stage Phase Lifecycle: [APPLY] → [EXECUTE] → [VERIFY] → [FINISH]

When virtual memory is active, work is divided into bite-sized phases. Every single phase proceeds through four rigorous stages:

The In-Chat Live Status Dashboard

During active phased execution, the agent must prepend this live status widget at the very top of every single response, giving you complete clarity on project state:

Automatic Cleanup & Archival Lifecycle

When the final phase passes all verification criteria:

1. The active main-plan.md and all phase-*.md files are moved to .agents/memory/archive/[Date]-[TaskName]/.
2. The active .agents/memory/phases/ folder is completely cleared.
3. The workspace memory is pristine and ready for the next task without token bloat.

Category Deep-Dives: Frontend & Anti-AI SOP

The frontend subsystem is engineered around one core benchmark: award-grade craft with zero AI tells and universal dynamic responsiveness. It combines the 247-site empirical study from Taste Kill v2 with modern typography, precision vector drafting, and 16.6ms frame-budget animation.

The Frontend Skill Suite

The Anti-AI SOP Quality Standard

The Anti-AI SOP suite acts as a ruthless filter across all text, code, vectors, and UI layouts:

Category Deep-Dives: Backend & Debugging

The Backend Subsystem: Karpathy Simplicity & TDD

The backend subsystem is built on Andrej Karpathy's core observation: LLMs tend to overcomplicate code by adding speculative flexibility, unnecessary abstractions, and defensive bloat. Skill Brain enforces minimal blast radius and strict test-driven development.

The Debugging Subsystem: Planning Gate & Evidence

Debugging in Skill Brain follows an uncompromising Iron Law: You are strictly forbidden from modifying code until the root cause is isolated with verifiable evidence.

Category Deep-Dive: Git, Commits & Branch Lifecycle

A core foundation of Skill Brain is that Git history is a first-class engineering deliverable. Monolithic, ambiguous, or AI-sloppy commits ("update files", "misc improvements") are strictly prohibited. The Git subsystem enforces automated multi-commit splitting based on domain blast radius, strict branch naming hygiene, and automatic Virtual Memory phase synchronization.

The Git Master Subsystem

The Multi-Commit Splitting Engine (`git-atomic-committer`)

When changes span multiple domains or exceed change size thresholds, Skill Brain automatically splits the changeset into 2 or more cohesive atomic commits:

Virtual Memory Phase Commit Integration (`git-phase-committer`)

Whenever a phase finishes in .agents/memory/phases/, the agent triggers git-phase-committer before proceeding. It commits application code, test fixtures, and memory markdown files in a structured 3-step sequence, embedding the exact CLI verification evidence into the git record.

Real-World Operations Walkthrough

Scenario 1: Dedicated Pre-Build UI Planning

User Prompt: /plan-frontend build a responsive analytics dashboard and metrics stream.

Scenario 2: Diagnosing an Intermittent Database Timeout

User Prompt: /debug our checkout endpoint intermittently throws 504 Gateway Timeout under moderate load.

Scenario 3: Adding a Multi-Phase Feature with Virtual Memory

User Prompt: /memory we need to build an OAuth2 Google and GitHub authentication system with role-based permissions.

Complete Skill Catalog Index (54 Skills)