By flight505
PRD-driven parallel development — generates PRDs, orchestrates Agent Teams for parallel story implementation with shared task coordination, two-stage review, and TDD enforcement.
Reviews full feature branch diff for code quality: architecture, security, types, tests, performance. Stage 2 of review — only runs after reviewer agent approves spec compliance. Returns structured verdict with file:line references.
Implements user stories from a shared task list. Claims tasks, codes with TDD discipline, runs quality checks, commits, and shares patterns with teammates. Designed for Agent Teams parallel execution.
Runs after all stories complete. Two-phase review of the full feature branch diff: spec compliance (verify each acceptance criterion with file:line evidence, check scope creep) then validation (run test/build/typecheck, verify commits).
Convert markdown PRDs into prd.json execution format for SDK Bridge autonomous agents. Use when a PRD document exists and needs to be converted to SDK Bridge's JSON format. Trigger on 'convert prd', 'convert to prd.json', 'turn this PRD into sdk-bridge format', 'create prd.json from this PRD'. Also used internally by /sdk-bridge:start at the conversion checkpoint. Do NOT trigger on general JSON conversion or plan execution requests.
Generate a Product Requirements Document (PRD) for SDK Bridge autonomous execution. Use when the user explicitly asks to create a PRD, write requirements, generate user stories for sdk-bridge, or scope a feature into a PRD. Also trigger on 'create prd', 'write prd for', 'spec out for sdk-bridge', 'generate requirements document'. Do NOT trigger on general planning, brainstorming, or task decomposition — those belong to other skills.
Uses power tools
Uses Bash, Write, or Edit tools
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.

PRD-driven parallel development. SDK Bridge turns your feature description into a structured PRD, then orchestrates multiple Claude Code Agent Teams teammates to implement stories simultaneously — with TDD discipline, quality gates, and two-stage review baked in.
One command. Parallel execution. No babysitting.
Based on Geoffrey Huntley's Ralph pattern.
You: /sdk-bridge:start
↓
Wizard: Ask 3-5 questions → generate tasks/prd-feature.md
↓
Review: Approve, improve, or edit the PRD
↓
Convert: prd.json + dependency graph analysis
"Group 1: US-001 (no deps)
Group 2: US-002, US-003 (parallel — both depend on US-001)
Group 3: US-004 (depends on US-002 + US-003)
Suggested teammates: 2"
↓
Configure: quality commands (test/build/typecheck) + code review
↓
Execute:
git checkout -b sdk-bridge/your-feature
TaskCreate × N stories
┌─────────────────────────────────────────────────────┐
│ Teammate 1 Teammate 2 Teammate N │
│ claim US-001 claim US-002 claim US-003│
│ TDD → commit TDD → commit TDD → commit│
│ TaskUpdate(done) TaskUpdate(done) TaskUpdate │
└─────────────────────────────────────────────────────┘
validate-task.sh fires on each TaskCompleted
check-idle.sh blocks any teammate from stopping early
↓
Review: reviewer subagent → full branch diff vs all criteria
code-reviewer subagent → architecture, security, types
↓
Done: prd.json marked complete · progress.jsonl updated · report
| Requirement | Version | Install |
|---|---|---|
| Claude Code CLI | v2.1.84+ | See Claude Code docs |
| Agent Teams | Experimental | export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 |
jq | Any | brew install jq (macOS) or apt install jq |
| Git | Any | Required for branch management |
Permissions: For autonomous execution (no permission prompts), start Claude Code with one of:
# Team/Enterprise plans — safer, uses background classifier
claude --permission-mode auto
# Max/Pro plans — skips all permission checks
claude --dangerously-skip-permissions
Without either, teammates inherit your default permission mode and will prompt for every file edit, bash command, and git operation. The permissionMode declared in agent frontmatter is overridden for Agent Teams teammates. Auto mode requires Team/Enterprise plan + Sonnet 4.6 or Opus 4.6.
Authentication (choose one):
# OAuth (recommended for Max subscribers)
claude setup-token
export CLAUDE_CODE_OAUTH_TOKEN='your-token'
# API key
export ANTHROPIC_API_KEY='your-key'
# Add the marketplace
/plugin marketplace add flight505/flight505-marketplace
# Install SDK Bridge
/plugin install sdk-bridge@flight505-marketplace
# Enable Agent Teams (add to ~/.zshrc or ~/.bashrc)
export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
# Start the wizard
/sdk-bridge:start
The wizard walks you through 5 checkpoints (dependency check → PRD generation → review → conversion → configuration), then takes over as team lead and runs Agent Teams until all stories are complete.

| Component | Count | Description |
|---|---|---|
| Commands | 1 | /sdk-bridge:start — interactive wizard + team lead |
| Skills | 2 | PRD generator, PRD converter (with dependency graph) |
| Agents | 3 | implementer (teammate), reviewer, code-reviewer |
| Hooks | 5 | TaskCreated, TaskCompleted, TeammateIdle, SessionStart, PreCompact |
Implementer (teammate, runs in parallel)
TaskList / TaskUpdateprogress.jsonl to share knowledge with other teammatesvalidate-task.sh hook blocks on test/build/typecheck failureReviewer (subagent, runs once after all stories complete)
git diff main...HEAD across all commitsnpx claudepluginhub joshuarweaver/cascade-ai-ml-agents-misc-2 --plugin flight505-sdk-bridgeFeature 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.
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.
Orchestrate multi-agent teams for parallel code review, hypothesis-driven debugging, and coordinated feature development using Claude Code's Agent Teams
Complete SOTA 2026 Storybook assistant with Vision AI design-to-code, natural language generation, AI-powered accessibility remediation, React Server Components, AI visual regression testing, design token sync, usage analytics, dark mode generation, and comprehensive testing (Storybook 10, React 19, Next.js 15)
Comprehensive project planning and architecture research skills for Claude Code - generates software architecture documents, sprint plans, building blocks, service cost analysis, and implementation roadmaps backed by real research.
AI-powered image, diagram, and video generation for Claude Code - uses Nano Banana 2 (Gemini 3.1 Flash Image) for fast generation, Nano Banana Pro (Gemini 3 Pro Image) for professional diagrams, and Veo 3.1 for video generation.
Long-running agent loop for Claude Code, in the Ralph pattern — a stable prompt, a mutable plan, and a loop that runs until the plan is empty.
Helper plugin for Claude Code's autonomous primitives — /goal, /branch, worktrees, dynamic workflows. Wraps native features with opinionated defaults and safety guardrails.