depot
DM-013/WORKS -- Design Machines' AI plugin marketplace for Claude Code and Codex.
Install
/plugin marketplace add Design-Machines-Studio/depot
Then install individual plugins:
/plugin install ned@depot
/plugin install craft-developer@depot
/plugin install project-manager@depot
/plugin install council@depot
/plugin install design-machines@depot
/plugin install assembly@depot
/plugin install live-wires@depot
/plugin install ghostwriter@depot
/plugin install design-practice@depot
/plugin install project-scaffolder@depot
/plugin install accessibility-compliance@depot
/plugin install dm-review@depot
/plugin install the-local@depot
/plugin install chef@depot
/plugin install pipeline@depot
/plugin install workflow-kernel@depot
/plugin install openrouter@depot
/plugin install airlift@depot
Codex compatibility is generated from the Claude manifests. Claude remains the source of truth; do not hand-edit Codex manifest files.
./tools/generate-codex-manifests.py
./tools/generate-codex-command-skills.py
./tools/validate-dual-compat.sh
Codex exposes Claude slash commands through generated skill aliases, so workflows
like /pipeline, /pipeline-run, /dm-review, and /dm-review-fix are also
available as pipeline:pipeline, pipeline:pipeline-run, dm-review:dm-review,
and dm-review:dm-review-fix.
Plugins
18 plugins | 39 domain-facing skills + 1 internal workflow-kernel skill + 34 generated Codex command-skill aliases | 38 agents | 34 commands
ned
Personal knowledge graph and session capture. 2 skills, 2 commands.
- ai-memory -- Interface to the ai-memory knowledge graph (~5,850 entities)
- recorder -- Captures conversation sessions into structured summaries, pushes to ai-memory and Notion
/capture-- Quick observation capture to ai-memory/depot-metrics-- Track and report plugin usage metrics via ai-memory
craft-developer
Craft CMS 4/5 development expertise. 4 skills, 3 agents, 4 commands.
- craft-development -- Element queries, Matrix fields, GraphQL, eager loading, debugging patterns, and a 30+ query cookbook
- content-modeling -- Three-layer content modeling methodology with Matrix patterns and CKEditor configuration
- craft-5-migration -- Craft 4 to 5 migration patterns, breaking changes, and upgrade checklists
- craft-mcp -- Craft CMS MCP server integration for direct database and schema access
- content-modeler (agent) -- Full content modeling workflow with MCP integration
- template-builder (agent) -- Template development with performance optimization
- craft-debugger (agent) -- Deep debugging with MCP tools and systematic analysis
/craft-debug-- Debug Craft CMS issues/craft-migrate-- Plan and execute Craft CMS migrations/craft-model-- Start a content modeling session/craft-query-- Build complex element queries
project-manager
LT10 methodology and Notion-integrated sprint planning. 2 skills, 1 command.
- lt10 -- Estimation (90th percentile), scoping, pricing, scheduling, risk management, stakeholder alignment, and communication scripts
- planner -- Notion-integrated sprint planning with Userback triage, Calendar.app meeting prep, Mail.app scanning, content ideation, and velocity tracking
/sprint-plan-- Run the full sprint planning workflow
council
Worker cooperative governance and decolonial content strategy. 2 skills, 1 agent, 1 command.
- governance -- BC Cooperative Association Act compliance, bylaw analysis, discovery processes, Co-op OS system design, voting thresholds, financial governance, and measurable membership-requirement modeling (track progress toward a target, not a boolean done-flag -- the lesson of Assembly Baseplate issue #233)
- decolonial-language -- Decolonial and anti-capitalist content strategy with terminology mappings between legal/regulatory language and solidarity economy alternatives
- governance-domain (agent) -- Reviews and guides cooperative governance feature development
/governance-check-- Quick cooperative compliance check
design-machines
Design Machines business strategy and product intelligence. 1 skill.
- strategy -- Product positioning, catalog system, Assembly architecture, partnerships, revenue model, brand language, design system (color palette, GT Standard typography, token architecture), and go-to-market strategy
assembly
Assembly governance application development with Go, Templ, and Datastar. 3 skills, 4 agents, 1 command.
- development -- Docker-based Go development with Templ templates, Datastar reactivity, page-type workflows, component library, database patterns, governance state machines, and federation trust choreography (mutual-trust handshake, TOFU pinning, key-rotation detection -- the open threat-model work behind Baseplate PR #252)
- nats-jetstream -- Embedded NATS patterns: KV store, event bus, streams, consumers, and real-time SSE via KV Watch
- golang-patterns -- SQLite, goose migrations, sessions, CSRF, rate limiting, cobra CLI, and CI security scanner diagnosis (gosec and gitleaks event modes, narrow false-positive remediation)
- templ-scaffolder (agent) -- Scaffolds new Templ pages, handlers, routes, and SSE endpoints
- datastar-sse (agent) -- Datastar reactivity and SSE endpoint patterns
- nats-reviewer (agent) -- Reviews NATS usage for embedded safety, ScopedEventBus patterns, and event ordering
- go-test-runner (agent) -- Runs profile-selected focused, level, candidate, and authenticated remote verification
/assembly-build-- Backward-compatible Docker build/test modes plus opt-in tiered repository verification
live-wires
Live Wires CSS framework for editorial websites. 1 skill, 1 agent.
- livewires -- Layout primitives (stack, grid, cluster, sidebar, cover, reel), baseline rhythm system, design tokens, cascade layers, container queries, and theming
- css-reviewer (agent) -- Reviews CSS and HTML for Live Wires framework compliance
ghostwriter
Travis Gertz's personal writing voice and editorial style engine. 2 skills, 2 agents, 1 command.
- voice -- Writing style covering tone, sentence rhythm, argumentation structure, vocabulary, anti-AI-writing patterns, and context-aware register switching
- social-media -- Platform-native social media strategy for LinkedIn, Instagram, Bluesky, and Mastodon
- voice-editor (agent) -- Reviews and edits content to match Travis's voice
- social-publisher (agent) -- Drafts platform-native social media content
/voice-check-- Quick text check against Travis's writing voice
design-practice
Design philosophy rooted in modernist masters and editorial tradition. 4 skills, 2 agents.
- typography -- From Muller-Brockmann, Gerstner, Bringhurst. Modular scales, vertical rhythm, typeface selection
- layout -- Editorial layout and art direction. Grid systems, visual hierarchy, pacing
- dataviz -- From Tufte, Wong, Franchi. Chart type selection, data-ink ratio, graphical integrity
- identity -- From Rand, Bass, Draplin, Wyman, Glaser. Bold simplicity, systematic extensibility
- design-critic (agent) -- Evaluates design work against the combined framework
- design-advisor (agent) -- Design decisions and strategic direction
project-scaffolder
Claude Code project infrastructure scaffolding. 1 skill.
- scaffolding -- Generates hooks, agents, settings.json, and CLAUDE.md templates for Go+Templ+Datastar, Go library, CSS framework, and Craft CMS projects
accessibility-compliance
WCAG 2.2 accessibility auditing and enforcement. 2 skills, 5 agents, 1 command.
- wcag-audit-patterns -- Automated and manual accessibility auditing for Live Wires, Go+Templ+Datastar, and Craft CMS
- screen-reader-testing -- VoiceOver, NVDA, and JAWS testing protocols with manual verification checklists
- a11y-html-reviewer (agent) -- Reviews HTML, Templ, and Twig templates for WCAG violations
- a11y-css-reviewer (agent) -- Reviews CSS for visual accessibility compliance
- a11y-dynamic-content-reviewer (agent) -- Reviews Datastar interactions and DOM updates for a11y
- a11y-audit-runner (agent) -- Runs automated audits using Pa11y, axe-core, and Playwright
- a11y-project-setup (agent) -- Sets up accessibility testing infrastructure
/a11y-audit-- Run automated accessibility audit on a page or template
dm-review
Code review orchestrator with parallel agents and deterministic repository quality pulses. 3 skills, 15 agents, 6 commands.
- review -- Single-command reviews launching up to 15 parallel agents tailored to Go+Templ+Datastar, Craft CMS, and Live Wires, with issue tracking, reorg-only PR verification for behavior-preserving file decomposition, and closure reconciliation across PR bodies, issue state, and review-finding issues
- visual-test -- Standalone visual browser testing for rendered pages using Playwright
- quality-pulse -- Scheduled/local repository audit using trusted profiles, authoritative JSON, explicit lane evidence, and compatible trends
- 12 review agents: security-auditor, architecture-reviewer, code-simplicity-reviewer, pattern-recognition-specialist, doc-sync-reviewer, go-build-verifier, craft-reviewer, test-coverage-reviewer, visual-browser-tester, ux-quality-reviewer, ui-standards-reviewer, migration-validator
- 1 second-opinion review agent: codex-perspective
- 2 workflow agents: review-consolidator, review-memory-recorder
/dm-review-- Full review with all applicable agents/dm-review-quick-- Quick review with 5 core criteria (6 logical lanes when OpenRouter adds its security lens)/dm-review-fix-- Resolve pending review findings from todos//dm-review-visual-- Visual browser testing on rendered pages/dm-review-loop-- Review-fix convergence loop until zero findings/dm-review-quality-pulse-- Run the repository quality-pulse workflow
the-local
Self-hosted Matrix network for the workplace democracy movement. 3 skills, 3 commands.
- element-branding -- Element Web custom CSS injection, welcome page customization, auth page styling
- synapse-config -- Synapse homeserver configuration, email setup via Resend SMTP, Jinja2 template system
- server-ops -- Docker Compose service management, SSH access, user creation, backups, deployment
/create-user-- Create a new Matrix user account/deploy-- Deploy changed files to production/logs-- Tail Docker Compose logs for services
chef
Science-driven cooking assistant with dietary analysis. 1 skill, 1 agent, 4 commands.
- cooking -- Recipes, dietary analysis (Eve Persak framework), Mela integration, Bali ingredient sourcing, food science
- recipe-analyzer (agent) -- Analyzes recipes against the dietary framework
/recipe-check-- Analyze a recipe against dietary guidelines/recipe-convert-- Convert a recipe to a healthy, Bali-friendly version/meal-plan-- Generate a meal plan following Eve Persak's timing framework/shop-- Build a shopping list organized by Bali store
pipeline
Autonomous feature development pipeline. 4 skills, 2 agents, 5 commands.
- assess -- Pre-plan assessment of current codebase state and UX before planning changes
- research -- Multi-source context gathering from ai-memory, RAG, web search, and domain plugins
- promptcraft -- Generates self-contained execution prompts with overlap-aware dependency ordering
- eval-sweep -- Ledger-first system-wide route, breakpoint, and accessibility evaluation
- plan-adversary (agent) -- Adversarial review of plans and prompts, iterating to convergence
- execution-orchestrator (agent) -- Autonomous worktree execution with batched review fixes, tiered tests, authenticated receipt reuse, and zero-deferral policy
/pipeline-- Full autonomous pipeline: assess, research, plan, prompt, review, execute, deliver/pipeline-assess-- Pre-plan assessment of current state/pipeline-prompts-- Generate execution prompts from an existing plan/pipeline-run-- Execute prompts in worktrees with review-fix loops/pipeline-fix-- Execute a bounded fix pass over numbered findings
workflow-kernel
Neutral deterministic workflow and inspection mechanics. 1 internal skill.
- workflow-kernel -- Run-state replay/recovery, behavioral contracts, authenticated tiered repository verification, receipts, trusted inspection profiles, contained lanes, redaction, canonical output, and compatible trends
openrouter
Unified external-model provider for quality-first agentic work and lower-cost mechanical review. 1 skill, 2 agents, 1 command.
- openrouter-delegate -- Threat-classified delegation to permitted open models for non-secret work, including security-related content when output and execution controls remain active
- openrouter-bulk-analyst (agent) -- Reviews eligible full diffs with quality-first Kimi K3 and a GLM-5.2 capacity fallback through the generic OpenRouter runner
- openrouter-agent-runner (agent) -- Runs dm-review mechanical agents through OpenRouter model slugs
/openrouter-- Delegate a task to any configured OpenRouter model
OpenAI and Anthropic models are prohibited through OpenRouter; use their native Codex and Claude CLIs.
airlift
Cross-harness session handoff. 1 skill, 3 commands.
- airlift -- Deterministic checkpoint and resume bundles across supported harnesses
/airlift-out-- Create or refresh a handoff bundle/airlift-in-- Resume from a handoff bundle/airlift-install-- Manage the optional early-warning monitor
Orchestration
Plugins compose through five patterns:
- Companion Skill Loading -- A command loads skills from other plugins at specific workflow phases
- Multi-Agent Dispatch -- A skill launches agents in parallel and consolidates results
- Memory-Mediated Coordination -- Plugins write to ai-memory entities that other plugins read later
- Pipeline Orchestration -- A conductor plugin composes all three patterns into an autonomous workflow with review-fix loops
- CLI-Mediated Model Delegation -- A plugin invokes an external model CLI/API and returns structured findings to the calling workflow
See docs/orchestration-patterns.md for details.
Validation
./tools/validate-composition.sh --all # run all validators
./tools/validate-dual-compat.sh # check Claude/Codex manifest sync and cache fallbacks
./tools/validate-quality-pulse.sh # run the 49-case offline pulse conformance suite
./tools/validate-marketplace-capabilities.sh # prove Claude/Codex discovery mappings
./tools/eval-descriptions.sh # run description trigger evals
./tools/check-dependencies.sh # check plugin dependencies
See Repository Quality Pulse and the Baseplate post-release handoff.
License
MIT