By OthmanAdi
Enforce a design-first development workflow using DESIGN.md plans, with automated guardrails that run on session start, stop, user prompts, and file writes. Block unauthorized code changes and automatically harvest code into DESIGN.md for continuous critique and memory routing.
One-shot wiring check for MDDesign. Verifies planning-with-files, DESIGN.md, code-memory-router, and Google linter are all reachable. Surfaces any missing piece with a one-line fix-it. Read-only.
Audit DESIGN.md against the actual code. Writes DESIGN.md.critique.md adjacent to DESIGN.md with five audit passes structural / drift / accessibility / completeness / consistency. Each finding has a stable id and a /mddesign:fix suggestion.
Deep diagnostic. Checks every MDDesign dependency, verifies hooks fire, validates the lint toolchain, tests scratch directory write, and prints a fix-it for every failure. Use when something is broken or before publishing to confirm a clean install.
Dispatch a phase to a specialist subagent via PhaseSpec v1. Auto-picks the subagent if not specified. HITL gated before dispatch. Result appended to progress.md under "Phase <id> Result".
Apply a single safe quick win from DESIGN.md.critique.md. HITL gated. Re-lints DESIGN.md after every edit; reverts if lint regresses.
Long-running codebase scan. Extracts de-facto design tokens from existing UI code, clusters them by frequency, names them, returns a draft DESIGN.md body conforming to the Google Labs spec. Dispatched by /mddesign:harvest.
Consumes one PhaseSpec v1 and runs it end to end. Reads referenced files itself. Returns a structured result. Default subagent for /mddesign:team dispatch.
Triggered on Stop. Classifies recent session output into scratch / WHERE / WHY tiers. Returns a list of WHY-promotion candidates with verbatim quotes and proposed MemPalace drawer paths. HITL gates the actual promotion.
Drafts a PhaseSpec v1 from a phase in task_plan.md. Read-only on planning files; returns the spec as text.
HITL surrogate. Audits an executor's structured result against the PhaseSpec done_when criteria. Returns approve, reject, or escalate. Read-only.
Continuously bridges DESIGN.md into the active planning phase. When the user is working on a UI phase, reads the project's DESIGN.md, extracts the relevant token slice, and appends it to findings.md under "## Design Context" so the agent sees design constraints in its attention window. Never edits DESIGN.md, never edits task_plan.md. Idles when DESIGN.md is absent.
Harvests a DESIGN.md from an existing codebase, critiques it against the actual code, and applies safe quick-win fixes. Three modes harvest / critique / fix. Shells out to `npx @google/design.md lint` for structural validation; never reimplements Google's linter. The on-ramp that makes DESIGN.md adoptable for every existing repo, not just greenfield projects.
Three-tier memory layer for the orchestrator. Routes scratch (file-tier), WHERE (QMD), and WHY (MemPalace) through the existing code-memory-router slash command. Never edits planning files directly. Every WHY-tier promotion goes through the HITL gate.
Makes sessions durable across `/clear`, machine reboots, and IDE switches. SessionStart catchup pulls scratch + last MemPalace recall + planning files into context. Stop promotion writes a handoff note and runs memory-keeper. Manual `/mddesign:handoff` re-runs either at any time.
Turns a plan phase into a PhaseSpec v1 and dispatches it to a specialist subagent via the Agent tool. Subagent reads files itself, returns a structured result, the result is logged into progress.md. Markdown is the shared state; no in-memory shared store across agents.
Modifies files
Hook triggers on file write and edit operations
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Glue layer for AI coding agents. Composes planning-with-files, Google's DESIGN.md, memory routing, agent teams, and session handoff into one orchestrator. Cross-IDE. No replacements.
You already use planning-with-files. You already use DESIGN.md (or want to). You already manage memory and agent dispatch by hand. MDDesign is the missing layer that wires them together so your agent sees the right design tokens at the right moment, dispatches the right subagent for the right phase, and never loses context after /clear.
It does not replace any underlying tool. It composes them.
| Command | What it does |
|---|---|
/mddesign:setup | Detect missing dependencies, print exact install steps for your platform |
/mddesign:doctor | Six-probe diagnostic. Use when something is broken |
/mddesign:compose | One-shot wiring check at the start of a project |
/mddesign:harvest | Scan your codebase, write a structurally valid DESIGN.md at project root |
/mddesign:critique | Five-pass audit: structural / drift / a11y / completeness / consistency |
/mddesign:fix <id> | Apply one finding from the critique. HITL gated. Re-lints after every edit. |
/mddesign:inject | Manually inject DESIGN.md tokens into the active phase's findings.md |
/mddesign:handoff | Force a session catchup or memory promotion. Survives /clear. |
/mddesign:memory | Three-tier memory: scratch / WHERE (QMD) / WHY (MemPalace) |
/mddesign:team | Dispatch a phase to a specialist subagent via PhaseSpec v1 |
DESIGN.md injected into findings.md. No more agents inventing hex codes./mddesign:harvest scans your existing code and writes one. The on-ramp for every existing project, not just greenfield./clear. Stop hook stamps a handoff. SessionStart hook reads the last stamp + the most recent decisions from MemPalace and rebuilds context in one paragraph.design-archeologist for harvest, a memory-keeper for classification, an executor for the rest. PhaseSpec v1 is the only handoff format.# Add the marketplace
/plugin marketplace add OthmanAdi/MDDesign
# Install the plugin
/plugin install mddesign@mddesign
/mddesign:setup
This walks you through every missing dependency with exact install steps. Safe to run multiple times.
Required:
npx (for @google/design.md lint)planning-with-files skillOptional but recommended:
code-memory-router skill (powers WHY/WHERE tiers)# In any project
/mddesign:compose # check the wiring
/mddesign:harvest # generate DESIGN.md from your existing code
/mddesign:critique # audit it for drift, a11y, completeness
That is the whole inner loop.
your existing code
│
▼
/mddesign:harvest ──► DESIGN.md (Google Labs spec, lint-validated)
│
▼
/mddesign:critique ──► DESIGN.md.critique.md
│
▼
on every UI phase, design-bridge auto-injects the relevant token slice
into findings.md "## Design Context" — your agent reads it on every turn
│
▼
/mddesign:team dispatch ──► specialist subagent (PhaseSpec v1)
│
▼
/mddesign:memory scratch ◄── session ends, Stop hook fires
WHERE ◄── QMD via code-memory-router
WHY ◄── MemPalace, HITL gated
│
▼
/mddesign:handoff ── SessionStart catchup pulls last
decisions back into context
Full architecture: docs/architecture.md.
These are load-bearing. MDDesign exists because it does not violate them.
npx claudepluginhub othmanadi/mddesign --plugin mddesignPersistent file-based planning for AI coding agents. Crash-proof markdown plans (task_plan.md, findings.md, progress.md) that survive context loss and /clear, with an opt-in completion gate and multi-agent shared state. Manus-style. Works with Claude Code, Codex CLI, Cursor, Kiro, OpenCode and 60+ agents via the SKILL.md standard. Includes Arabic, German, Spanish, and Chinese (Simplified and Traditional).
Build generative UI with OpenUI — any LLM provider, any backend language. Scaffold, integrate, validate.
HTML-first persistent planning skill. Single self-contained plan.html with interactive phases, drag-and-drop, sliders, mockups, copy-as-markdown export, embedded JSON state. Survives /clear via session catchup. Tamper-protected by SHA-256. Mirrors across 17 IDEs.
Build autonomous agents in Rust, TypeScript, or Python using Open Responses API for seamless multi-provider support (OpenAI, Anthropic, HuggingFace, etc.)
AI text humanizer skills for Arabic and Hebrew — MSA, Egyptian, Levantine (Syrian/Lebanese/Palestinian), and Modern Hebrew.
Feature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Production-grade engineering skills for AI coding agents — covering the full software development lifecycle from spec to ship.
Persistent file-based planning for AI coding agents. Crash-proof markdown plans (task_plan.md, findings.md, progress.md) that survive context loss and /clear, with an opt-in completion gate and multi-agent shared state. Manus-style. Works with Claude Code, Codex CLI, Cursor, Kiro, OpenCode and 60+ agents via the SKILL.md standard. Includes Arabic, German, Spanish, and Chinese (Simplified and Traditional).
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Complete developer toolkit for Claude Code
Intelligent draw.io diagramming plugin with AI-powered diagram generation, multi-platform embedding (GitHub, Confluence, Azure DevOps, Notion, Teams, Harness), conditional formatting, live data binding, and MCP server integration for programmatic diagram creation and management.