jb-612

DDD 7-Layer scaffold before TDD. Transforms approved user stories, acceptance criteria, functional design, and feature task breakdown into a production-grade, layered code skeleton — folders, interfaces, aggregates, value objects, domain events, test…

3-round multi-tier design review with 6 specialist reviewers (including HLD compliance), adversary challenge, cross-debate, and synthesis. Use during design pipeline stages 5/7 or standalone before TDD build.

Run design + review + TDD for 2-4 coupled features concurrently via native teams. Each feature gets a dedicated fresh-context designer + runner; team-lead brokers coupling decisions in real-time and runs a consolidated cross-feature review before the user…

Clean-slate architecture research for brownfield modernization. Evaluates open-source alternatives, design patterns, and build-vs-adopt decisions using multi-agent research teams with adversary debate. Part of the Brownfield Refactoring Toolkit.

Multi-perspective research with parallel agents — web research, codebase analysis, graph queries, and cost/risk analysis — synthesized into a structured report for design decisions.

11-stage pipeline — HLD-driven design, PRD-driven stories via meeting room brainstorm, 2-round review, HLD-traced task breakdown, user approval. ACM graph-compatible traceability.

Cross-module boundary tests between layers

Lightweight refactoring design skill — scaffolds workitems, fills design.md and tasks.md from existing research, single HITL gate. Skips user stories and multi-round review since behavior is already pinned by characterization tests.

Scaffold a new ADK function tool with TDD lifecycle and eval case addition

Adversarial testing loop inspired by autoresearch. Generate adversarial cases, run against agent, evaluate, accumulate findings.

Three-pass code review checking prerequisites, code quality, and test/eval coverage

Create a conventional commit with workitem traceability and pre-commit checks

Design conversation flows and state transitions for the mortgage concierge phases

Design and implement a new ADK sub-agent with tools, state contracts, and eval cases

Review a feature design for completeness, architecture alignment, and ADK patterns compliance

Deep analysis of ADK eval results with failure pattern identification and improvement suggestions

cursor-sim API contract reference. Use when implementing services that consume cursor-sim data (analytics-core, viz-spa), ensuring API alignment, or verifying response formats. Source of truth for data models and endpoint contracts. (project)

Cursor Business API implementation patterns. Use when implementing API endpoints, writing HTTP handlers, handling authentication, pagination, error responses, or CSV exports. Covers response formats for AI Code Tracking, Analytics, and Admin APIs. (project)

Identifies required updates to related files after primary changes. Use after completing any code changes to detect documentation drift, missing test updates, and cross-file dependencies. Triggers on refactoring, model changes, or interface modifications.…

Go coding standards for cursor-sim service. Use when writing Go code, implementing HTTP handlers, creating structs, handling errors, or writing Go tests. Covers naming conventions, error handling, concurrency, and testing patterns. (project)

Guide for selecting the right Claude model (Haiku, Sonnet, Opus) for different tasks. Use when starting a new task, planning implementation, or optimizing costs. Helps choose between fast/cheap Haiku and powerful Sonnet/Opus. (project)

React and Vite best practices for cursor-viz-spa dashboard. Use when implementing React components, Apollo Client queries, Tailwind styling, or Vite configuration. Covers hooks, state management, and data visualization patterns. (project)

Post-task completion checklist for Spec-Driven Development. Use after completing any implementation task to ensure proper commit hygiene. Triggers on "task complete", "step done", "ready to commit", or moving to next task. (project)

Create or revise technical design documents. Use when planning architecture, making component decisions, designing interfaces, or documenting ADRs (Architecture Decision Records). Produces design.md files with rationale and alternatives. (project)

Two-phase code review: automated quality gates then manual inspection for data integrity, API correctness, security, and types.

Conventional commits for Red Alert project. Enforces commit format and pre-commit validation.

Explore and validate OREF/Tzofar data sources. Fetch samples, check schemas, verify BOM handling.

Run comprehensive quality gates: tests with coverage, linting (25 rule sets), type checking, complexity, dead code. Use before commit.

Multi-perspective research with parallel agents — web research, codebase analysis, graph queries, and cost/risk analysis — synthesized into a structured report for design decisions.

Evaluate ROI of reinforcement learning and fine-tuning of foundation models via a structured 5-phase research project (parallel market+academic discovery → stream selection → parallel deep research → adversarial review → synthesis). Produces per-chapter…

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or…