Shafqat Ullah shafqatevo

Framelink export — homepage design context

1. Purpose

Export the homepage design from Figma so Shafqat can rebuild it offline with AI tools (Path A / local MCP). You only need Figma edit access and the Framelink Exporter plugin. You do not need WordPress, tokens, or any project tooling.

2. Target file and node

Item	Value

Figma → WordPress: AI Agent Prompt (FIGMA2WP)

Purpose: Executable system prompt for an AI coding agent. Converted from the canonical human+agent guide FIGMA2WP.md with no information loss.
Naming: Use Alumnium (alumnium.ai). If you see “Aluminum” in older notes, it means Alumnium.
Self-contained: No Silver Bullet / Sidekick. Path A is mandated. Path B is appendix-only.
Companions: Prefer on-disk qa/ files when present; if missing, write them from the inlined sources in this prompt before running.

SYSTEM / ROLE

Consolidated AI Report: Using OpenCode + MiniMax M3 + Figma MCP for Pixel-Perfect WordPress Implementation

Figma: hGzBbWvAx1jTNEPtgS3jDV / Node 222-1554 → Target: jobready.global Responses synthesized from: Gemini, Claude, ChatGPT, Copilot, Perplexity, Grok & DeepSeek

IMPORTANT CONTEXT: What OpenCode + MiniMax M3 + Figma MCP Actually Is

OpenCode is a terminal-based AI coding agent (similar to Claude Code/Cursor) that operates in your local dev environment. MiniMax M3 is a multimodal LLM with vision + deep reasoning that can be used as OpenCode's underlying model. Figma MCP (Model Context Protocol) is a standardized bridge that allows the AI agent to directly read Figma design data (layers, tokens, spacing, colors, typography) without manual export — this is the key to pixel-perfect output.

LeanCTX vs RTK + Context Mode + agentmemory + Graphify

Scope: Pure capability and feature comparison only — no Silver Bullet migration cost, hooks/gates enforcement policy, licensing, pricing, or adoption recommendations. Baseline: LeanCTX as a single product vs the RTK + Context Mode + agentmemory + Graphify composable stack. Evidence: ultradeep matrix audit (2026-07-05), 200 feature rows × 5 columns, 1000 cells verified (audit-report.md).

Pure capability replacement analysis

Verdict

LeanCTX is not a full replacement for the four-tool stack. It overlaps heavily on compression, sandbox analysis, hooks, memory, and code-graph surfaces, but the audited matrix and replacement pass show residual hard gaps (especially agentmemory orchestration and Context Mode fetch/hook depth) and many partial (¹) / composable (²) cells where LeanCTX matches intent yet not parity.

SonarSource LLM Comparison: GPT-5.5 Medium vs Claude Opus 4.7 High — Analysis Transcript

Generated 2026-06-02. This is a verbatim transcript of a conversation analyzing the SonarSource LLM code quality comparison leaderboard.

	{
	"$schema": "https://opencode.ai/config.json",

	"agent": {
	"ocg-minimax-m3": {
	"mode": "subagent",
	"model": "opencode-go/minimax-m3",
	"description": "OCG subagent on opencode-go/minimax-m3 (top-tier MiniMax M-series). General-purpose; pick this when the parent wants minimax-m3 specifically."
	},
	"ocg-qwen3.7-max": {

	{
	"$schema": "https://opencode.ai/config.json",
	"mcp": {
	"agentmemory": {
	"command": [
	"npx",
	"-y",
	"@agentmemory/mcp"
	],
	"environment": {