managing-design-briefs

Brief-before-generation system with locked design tokens and anti-slop enforcement for Loxihart and FlashFusion brand profiles. Use this skill whenever someone asks to build, spec, audit, or generate

SKILL.md

GitHub

Managing Design Briefs

Brief-before-generation with locked tokens, anti-slop gates, and adversarial critic. Two brand profiles (Loxihart, FlashFusion) share one workflow — tokens and aesthetic direction differ.

Evals

Should trigger:

  1. "I'm about to build a dashboard panel — let's start"
  2. "Does this design pass the anti-slop gates?"
  3. "Define the tokens before I generate this in v0"

Should NOT trigger:

  1. "Build the React component directly, skip the brief" — only acceptable when design direction is already fully specified; otherwise the brief is mandatory first (this skill owns that gate)
  2. "Write pure backend logic with no UI component" — no visual deliverable → route to a backend/build skill (e.g. kyle-vibe-coding)

Hard Rules

  1. Brief before generation. Always. No exceptions. If no brief exists — stop, create one, wait for confirmation.
  2. Profile confirmed before tokens load. Never mix profiles mid-session.
  3. All 7 anti-slop gates pass before any output. No partial passes. Fix or stop.
  4. Adversarial critic runs on every brief before generation. Not optional for STANDARD/FULL scope.

Quick-Start Decision Tree

Request arrives
  │
  ├─ Confirm profile first ─────────────────────────────────────────────
  │   ├─ Loxihart      (enterprise tool, internal dashboard, MSP context)
  │   └─ FlashFusion (product UI, landing page, artifact, app.ff.co)
  │
  ↓ (profile confirmed)
  │
  ├─ New brief needed ──────────────── workflows/brief-creation.md
  ├─ Audit existing brief ─────────── workflows/anti-slop-gates.md → gap list
  ├─ Adversarial critic pass ──────── workflows/adversarial-critic.md
  ├─ Extract generator prompt ──────── templates/generator-prompt-template.md
  ├─ Token drift check ────────────── workflows/token-drift-audit.md
  │                                   (or run scripts/token-drift-check.py)
  ├─ Update locked tokens ─────────── Edit references/[profile]-tokens.md
  │                                   Note version + reason
  ├─ Multi-session / 3+ screens ────── references/advanced.md
  └─ Sales / doc / SPIKE ──────────── Not UI → redirect

Profile Summary (Locked Tokens in references/)

Load only the active profile — never both.

Profile Token file Aesthetic
Loxihart references/Loxihart-tokens.md Flat void surface, electric violet, Space Grotesk + Inter, flat cards, dense enterprise
FlashFusion references/FlashFusion-tokens.md Void-dark #0B0B14, blur(12px) glass cards, Instrument Serif + Geist, violet/magenta/cyan

Cross-profile contamination is a hard failure. See each token file's isolation section.


Workflow Routing

Task Where to go
Creating a new brief workflows/brief-creation.md
Running anti-slop gates workflows/anti-slop-gates.md
Adversarial critic review workflows/adversarial-critic.md
Checking output for token drift workflows/token-drift-audit.md or scripts/token-drift-check.py
Filling the brief template templates/brief-template.md
Building the generator prompt templates/generator-prompt-template.md
OKLCH / multi-session patterns references/advanced.md

Scope Gate

  • LITE (single component, one-shot): gates 1–7 + brief template. Skip adversarial critic.
  • STANDARD (multi-component, dashboard, page): full workflow + adversarial critic.
  • FULL (design system, multi-session, pipeline): full workflow + adversarial critic + dual-file pattern from references/advanced.md.