Claude Skills

Suggests manual context compaction at logical intervals to preserve context through task phases rather than arbitrary auto-compaction.

Universal coding standards, best practices, and patterns for TypeScript, JavaScript, React, and Node.js development.

Frontend development patterns for React, Next.js, state management, performance optimization, and UI best practices.

Design database schemas, plan migrations, optimize queries, and manage data models. Covers relational databases (PostgreSQL, MySQL, SQLite), document databases (MongoDB), and ORM integration (Prisma, Drizzle, TypeORM). Use this skill when designing schemas, reviewing data models, planning migrations, optimizing slow queries, or establishing database patterns for a project. Triggers on "database", "schema", "migration", "model", "query optimization", "index", "normalize/denormalize".

Extended investigation for technology evaluation, competitive analysis, architectural exploration, and best practices research. Unlike spike (time-boxed hours), deep-research supports multi-day comprehensive investigation producing reports for major decisions. Triggers on "research X solutions", "compare frameworks for our use case", "evaluate vendors for X", "what are best practices for Y", "how do competitors handle Z", or any investigation requiring comprehensive analysis across multiple sources.

Implement logging, metrics, tracing, and alerting for production applications. Covers structured logging (Pino, Winston), metrics (Prometheus, DataDog, CloudWatch), distributed tracing (OpenTelemetry), and alert design. Use this skill when adding logging to services, setting up monitoring, creating alerts, debugging production issues, or designing SLIs/SLOs. Triggers on "logging", "monitoring", "alerting", "observability", "metrics", "tracing", "debug production", "correlation id", "structured logging", "dashboards", "SLI", "SLO".

Guided plan revision with impact analysis. Supports explore mode (discover what to change) and direct mode (apply known changes). Always shows impact before execution.

Create, extract, and publish TypeScript SDKs. Covers analyzing existing applications to extract reusable logic, designing clean SDK APIs, implementing typed clients with proper error handling, bundling for multiple targets (ESM/CJS/browser), and publishing to npm (public or private registries). Use this skill when building SDKs, extracting shared code into packages, or creating developer tooling libraries.

Primary SystemVerilog orchestrator. Handles all RTL tasks end-to-end - routes to agents, runs verification, iterates until working. Examples: "create a FIFO", "test my UART", "fix lint errors", "implement and verify"

Summarize Verilator/lint output in a readable format

Intent router and expand mode orchestrator for GateFlow. Classifies user intent semantically, determines confidence, triggers expand mode for ambiguous requests, and hands off to appropriate skill/agent.

SystemVerilog lint checker with structured output. Runs Verilator lint, categorizes errors/warnings, explains issues, returns parseable result block for orchestration.

SystemVerilog learning mode - generates exercises, reviews solutions

Expand mode - asks clarifying questions, presents options with trade-offs, then hands off to appropriate skill/agent with enriched context.

Parallel build orchestrator for SystemVerilog creation tasks. Decomposes designs into independent modules, builds in parallel phases, runs verification on each component, then integrates. Example: "/gf-build RISC-V CPU with ALU, regfile, decoder, control FSM"

Provide architecture-specific GPU optimization advice for NVIDIA GPUs (Ampere, Hopper, Ada). Use when discussing GPU architecture features, compute capability, or hardware-specific optimizations.

Optimize CUDA kernels and GPU code for performance. Use when reviewing CUDA code, analyzing performance, or suggesting GPU optimizations.

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup

Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies