Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Fullstack development agent for the A3 platform — orchestrates Ember.js, Firebase/Firestore, and GCP Cloud Functions with deep codebase knowledge, round-robin review, and GitHub-authenticated access control
npx claudepluginhub trusted-american/marketplace --plugin a3-pluginStandalone ability & security specialist — create ember-can abilities and Firestore rules for A3
Standalone Glimmer GTS component specialist — create, modify, or get expert advice on A3 components
Standalone TAIA design system specialist — use design system components, check compliance, and get advice on @trusted-american/ember usage in A3
Find real examples and established conventions within the A3 codebase — search by feature type, component pattern, or convention question
Standalone Cloud Function specialist — create Firebase/GCP Cloud Functions for A3
Specialist agent for writing ember-can ability files and Firestore security rules in the A3 application. Deep knowledge of A3's permission model, role-based access control, and how frontend abilities must align with backend Firestore rules. <example> Context: Permissions needed for a new referral feature user: "Create abilities for referrals — admins can CRUD, agents can only read their own" assistant: "I'll create the ability file in app/abilities/referral.ts following A3's base ability pattern, and update firestore.rules to match. Let me read the existing ability and rules patterns first." <commentary> The ability-writer ensures frontend abilities and backend Firestore rules are always in sync — a security requirement that the integration-specialist also validates. </commentary> </example>
Final quality gate agent for A3 code review. Has holistic knowledge of A3 conventions, security practices, performance patterns, and code quality standards. This agent has veto power in the round-robin review process and checks all code from every specialist. <example> Context: Round-robin review of a complete feature implementation user: "Review all code from the referral feature implementation" assistant: "I'll perform a comprehensive review across all files: conventions compliance, security, performance, TypeScript strictness, and A3 pattern adherence. Any issues will block acceptance." <commentary> The code-reviewer is the last line of defense. They check everything holistically and can veto even if all specialist agents approved. </commentary> </example>
Specialist agent for writing Glimmer GTS components in the A3 Ember.js application. Deep knowledge of Glimmer component patterns, tracked properties, GTS template syntax, Tailwind CSS + Bootstrap styling, and A3's component conventions. <example> Context: A new enrollment status badge component is needed user: "Create a badge component that shows enrollment status with color coding" assistant: "I'll create a Glimmer GTS component following A3's badge pattern in app/components/badges/. Let me first read the existing badge components to match conventions." <commentary> The component-writer reads existing A3 badge components to match naming, signature, and styling patterns before generating the new component. </commentary> </example> <example> Context: A complex form editor component is needed user: "Create a multi-step editor for group enrollment intake" assistant: "I'll build this as a GTS component in app/components/editors/ following A3's editor patterns. Let me trace the existing editor components to understand the state management and form validation patterns." <commentary> For complex components, the agent investigates existing editors, form patterns, and validation approaches before writing code. </commentary> </example>
Specialist agent for the @trusted-american/ember design system. Deep knowledge of all 88+ Glimmer GTS components, 4 helpers, 3 modifiers, design tokens, and the core Tailwind class system. Ensures all A3 UI code uses design system components wherever applicable and follows TAIA brand standards. <example> Context: A new feature needs a form with inputs, select, and a badge user: "Create a referral form using the design system components" assistant: "I'll build this using the design system's Form::Input, Form::Select, Badge, Button, and Card components from @trusted-american/ember. Let me check the component signatures first." <commentary> The design-system-writer knows every available component in the TAIA design system and will use them instead of raw HTML/Bootstrap. It checks the design-system repo for exact signatures. </commentary> </example> <example> Context: Reviewing another agent's component that uses raw HTML buttons user: "Review this component for design system compliance" assistant: "I see raw `<button class='btn btn-primary'>` usage — this should use `<Button @color='primary'>` from the design system. Also the form inputs should use `<Form::Input>` instead of raw `<input class='form-control'>`." <commentary> The design-system-writer catches opportunities to replace raw HTML/Bootstrap with proper design system components during round-robin review. </commentary> </example>
Codebase archaeologist agent that finds real examples and established conventions within the A3 repository. Searches across 1,600+ app files and 800+ tests to find how things are actually done — not how docs say they should be done. Used by other agents to ground their output in real A3 patterns, and during review to verify convention compliance. <example> Context: The component-writer needs to create a new badge component user: "Find examples of how badge components are structured in A3" assistant: "I found 48 badge components in app/components/badges/. Here are the 3 most relevant patterns: enrollment-status.gts uses a STATUS_COLORS map with @color arg, contract-status.gts uses computed getters for label text, and ticket-status.gts shows the simplest pattern with a direct switch. All use the design system Badge component." <commentary> The example-finder doesn't guess — it reads the actual codebase to find real, working examples that other agents should follow. This prevents convention drift. </commentary> </example> <example> Context: The route-writer is about to create a new list route with filtering user: "How do existing list routes with query params work in A3?" assistant: "I found 12 routes with queryParams controllers. The most common pattern is in authenticated/clients.ts + controllers/authenticated/clients.ts: queryParams=['search','status','page'], tracked properties with defaults, and resetPage action. The template passes these to a search-list component. 8 of 12 use this exact pattern." <commentary> The example-finder provides statistical confidence — "8 of 12 routes do it this way" — so other agents know which pattern is the established convention vs. an outlier. </commentary> </example> <example> Context: During round-robin review, checking if new code matches conventions user: "Does this new Firestore trigger follow A3's patterns?" assistant: "Comparing against 45 existing Firestore triggers: A3 triggers always (1) check for null snapshot, (2) wrap in try/catch with Sentry, (3) create an activity document for audit trail. Your trigger is missing the activity creation — every other onCreate trigger in A3 creates one. Here are 3 examples showing the pattern." <commentary> During review, the example-finder acts as a convention enforcer by showing concrete evidence of what the codebase actually does, with counts and specific file references. </commentary> </example>
Complete architecture reference for the A3 insurance platform — file locations, data flow, conventions, microservice map, and fullstack patterns
Algolia search integration reference — 3 backend files + frontend search service. Index management, record syncing, search queries, and API key management
A3 authentication and permission system — Firebase Auth, ember-simple-auth, ember-can abilities, role-based access, and the full auth flow
A3 build system reference — Embroider, Vite, Babel, TypeScript, route splitting, PWA support, and build configuration
Deep Google Cloud Functions v2 reference — every trigger type with full TypeScript signatures, function configuration, retry behavior, cold start optimization, testing, deployment, emulator usage, A3 index.ts export pattern, error handling, structured logging, all 40 Firestore trigger files by collection, and all 39 HTTPS endpoint files by service
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Google Firebase MCP integration. Manage Firestore databases, authentication, cloud functions, hosting, and storage. Build and manage your Firebase backend directly from your development workflow.
Use this agent when evaluating new development tools, frameworks, or services for the studio. This agent specializes in rapid tool assessment, comparative analysis, and making recommendations that align with the 6-day development cycle philosophy. Examples:\n\n<example>\nContext: Considering a new framework or library
Advanced plugin marketplace platform with intent-based composition, supply chain security, contextual intelligence, dev studio hot-reload, and federated registry protocol. Transforms the marketplace from a package manager into an enterprise orchestration platform.
Firebase platform expert for Firestore, Auth, Functions, and Vertex AI integration
Full-stack orchestration with deployment, performance, security, and test automation
Complete development suite: 3 expert agents (fullstack developer, validation gates, documentation manager) + 3 commands (containerize, PRP generation/execution) + 5 skills (git commit helper, webapp testing, devtools, PRP generator, Fifteen-Factor App) + 5 MCP integrations
Multi-agent algorithm intelligence system that analyzes codebases for optimization opportunities, researches cutting-edge algorithms via deep web research, identifies complexity bottlenecks, and applies superior data structures and algorithms with before/after benchmarking. 3 commands, 7 agents, 1 skill, 3 templates.
Multi-agent Playwright test generation system that creates thorough, production-ready test suites one page at a time
Multi-agent reverse engineering system with 10 specialized agents, 3 commands, and continuous fidelity assurance. Deeply researches any software via Firecrawl, Perplexity, and web intelligence; extracts architecture and features; analyzes competitors for UX, adoption, and feature parity; generates implementation blueprints; builds autonomous clones with orchestrated agent teams; and maintains fidelity through automated drift detection and self-healing loops. 3 commands, 10 agents, 1 skill, 4 templates.
Plugin registry for Claude Code, maintained by Trusted American Insurance Agency.
marketplace/
├── plugins/ # First-party plugins
├── community/ # Third-party forks and adaptations
├── .claude-plugin/
│ └── marketplace.json # Auto-generated plugin index (do not edit)
├── tools/
│ ├── lib/registry.js # Shared registry generation logic
│ ├── generate-marketplace-json.js # CLI script to regenerate marketplace.json
│ └── marketplace-mcp/ # MCP server for plugin management
├── .github/workflows/ci.yml # Validation + registry generation
└── .mcp.json # Auto-loads the MCP server in Claude Code
Every plugin in plugins/ or community/ must contain these files:
| File | Purpose |
|---|---|
README.md | Usage documentation |
LICENSE | License file |
.claude-plugin/plugin.json | Plugin manifest |
Components go in the plugin root, never inside .claude-plugin/:
plugin-name/
├── .claude-plugin/
│ └── plugin.json # Manifest — only file in this directory
├── skills/ # Each skill is a subdirectory with a SKILL.md
│ └── skill-name/
│ └── SKILL.md
├── agents/ # Agent definitions (.md files)
│ └── agent-name.md
├── commands/ # Slash command definitions (.md files)
│ └── command-name.md
├── hooks/ # Hook configurations (.json files)
│ └── hooks.json
├── templates/ # Output templates (.md files)
│ └── template-name.md
├── .mcp.json # MCP server config (optional)
├── README.md
└── LICENSE
The name field is required and must be kebab-case (^[a-z0-9-]+$). All other fields are optional but recommended.
{
"name": "my-plugin",
"version": "0.1.0",
"description": "What this plugin does",
"author": { "name": "Author Name" },
"license": "MIT",
"keywords": ["keyword1", "keyword2"],
"repository": "https://github.com/org/repo"
}
Names must be unique across the entire marketplace. If two plugins share a name, CI will warn and the second one overwrites the first in .claude-plugin/marketplace.json.
.claude-plugin/marketplace.json is a generated index of all plugins. It is not updated in PRs. The flow:
main.node tools/generate-marketplace-json.js, which scans plugins/ and community/, reads each plugin.json, collects component directories, and writes .claude-plugin/marketplace.json.main with [skip ci] to avoid a loop.The lastUpdated timestamp updates on every run. The generation logic lives in tools/lib/registry.js and is shared between the CLI script and the MCP server.
The MCP server at tools/marketplace-mcp/ auto-loads when you open this repo in Claude Code (configured in .mcp.json). It exposes 6 tools:
| Tool | What it does |
|---|---|
create_plugin | Scaffolds a new plugin with required files and correct directory structure |
validate_plugin | Checks a plugin for required files, valid manifest, and component structure |
validate_all | Validates every plugin in plugins/ and community/ |
list_plugins | Lists all plugins with version, description, and component breakdown |
add_component | Adds a skill, agent, command, hook, or template to an existing plugin |
get_conventions | Returns the full marketplace conventions document |
Validation checks: required files exist, plugin.json is valid JSON with a kebab-case name, and any component directories that exist contain correctly typed files (.md for skills/agents/commands/templates, .json for hooks).
Every push and PR runs two jobs:
plugins/ and community/ for README.md, LICENSE, and a valid .claude-plugin/plugin.json with a kebab-case name.npm test in tools/marketplace-mcp/ (vitest, 57 tests covering validation, registry generation, component handling, and security).On merge to main, a third job runs:
.claude-plugin/marketplace.json and commits it.Prerequisites: Claude Code and Node.js 18+ (22 recommended).
git clone https://github.com/trusted-american/marketplace.git
cd marketplace
npm ci --prefix tools/marketplace-mcp
Open the repo in Claude Code. The MCP server loads automatically. Run list plugins to verify.
If you use the jira-orchestrator plugin for Jira/Confluence integration: