Coding standards, frameworks, refactoring, and code review (18.6K skills)
Break a complex, multi-session engineering objective into a Markdown plan of executable steps. Each step includes its own context, tasks, checks, and completion criteria, with dependencies and parallel work identified and the plan reviewed for gaps.
Plan and improve server-side applications built with Node.js, Express, and Next.js API routes. Get practical patterns for API structure, database queries, caching, authentication, error handling, background jobs, and logging.
Supports writing, reviewing, and refactoring modern C++17, C++20, and C++23 code using the C++ Core Guidelines. It helps check type safety, resource ownership, error handling, concurrency, templates, naming, and standard library usage for clear, consistent code.
Query a CodeGraph map of an initialized codebase to find symbols, callers, callees, focused context, and likely change impact. Summarize the structural evidence for planning, implementation, or review workflows.
Structure Android and Kotlin Multiplatform apps with clear module boundaries and dependency rules. Apply UseCase and Repository patterns, organize data sources, and keep domain models separate from database and network details using examples for common Kotlin tools.
Coordinate coding work across Claude Code agents with Claude DevFleet. Turn a project request into linked missions, run agents in separate Git worktrees, track their progress, and review structured reports of changes, errors, and next steps.
Helps build applications with the Anthropic Claude API and SDKs in Python or TypeScript. It covers messages, streaming, tool use, image input, batch requests, prompt caching, error handling, and multi-step agent workflows, with examples for common integration tasks.
Use this skill to build shared Android, iOS, desktop, and web interfaces with Compose Multiplatform or Jetpack Compose. It provides practical patterns for UI state, navigation, reusable composables, theming, platform-specific UI, and reducing unnecessary recomposition.
Design or review REST APIs with consistent resource names, HTTP methods, status codes, response formats, pagination, filtering, rate limits, and versioning. The guidance helps shape endpoints that are clear for developers and suitable for public or internal services.
Choose Claude Code skills and rules to install at user or project level. The workflow checks installed files and path references, reports problems, and can tailor the installed materials to a project's tech stack and preferences.
Define or review an API contract so architects, frontend and backend developers, and QA can agree on requests, responses, authentication, errors, and compatibility. It helps teams record who calls an interface and what each side must know before implementation and verification.
Apply a file-processing cache keyed by SHA-256 content hashes. Results can be reused after a file is moved or renamed, while changed content naturally gets a new cache key. The pattern keeps caching in a service layer separate from pure processing functions.
Record significant architecture choices as concise ADR documents alongside the code. Capture the context, alternatives, rationale, and consequences so developers can understand why a framework, database, API, or other design direction was chosen.
Build an AI agent project through six collaborative stages, from requirements and architecture to tools, prompts, code generation, and verification. Start framework-independent, then generate a runnable project for frameworks such as LangChain, EINO, AutoGen, AgentScope, or CrewAI.
Assess a concrete long-running Python entrypoint to see where repeated work can safely run in parallel and whether the overhead is worthwhile. When source optimization is authorized, make a minimal refactor and require correctness and performance evidence before claiming an improvement.
Configure a CarsXE API key by validating it with the CarsXE API, saving it for future sessions, and setting it for the current session. It can also check an existing key without changing it.
Set engineering practices for teams where AI agents produce much of the code. Focus planning on clear acceptance criteria, review on system behavior and risk, and use deterministic tests and integration checks to evaluate generated changes.
Set up an app backend through Cohesivity’s HTTP API, including databases, hosting, authentication, storage, realtime features, and AI model access. The skill guides agents through project setup and service provisioning while handling credentials and user approval requirements.
Fetch available vehicle years, makes, models, trims, or variants for cascading dropdowns using the CarsXE YMM Options API. It presents the matching options and can continue to the next selection level; full vehicle specifications are handled separately.
Create, repair, audit, consolidate, or export agent skill packages built around SKILL.md. It checks for existing owners and evidence before proposing new skills, and can produce reports or platform-specific exported files for selected operations.
Takes one accepted change from implementation through required checks and closeout. It uses a fresh, author-distinct review only when requested or when a mistake could be costly, and reports evidence, unchecked work, and open decisions.
Get evidence-based advice on a plan, design, or code change. The skill identifies actionable findings with locations, consequences, and suggested next checks, while stating what it inspected and what remains unknown. Its feedback is advisory, not an acceptance verdict.
Restructure code to extract, simplify, rename, or remove a bounded piece while preserving what callers observe. Compare focused checks before and after, including relevant error paths and output, and report behavior that could not be checked.
Operate a Gas City software factory when it has been explicitly selected. Hand a task to its Mayor, inspect run, session, bead, and artifact state, and diagnose stalled workers. The skill returns runtime evidence; it does not decide whether the work is correct or complete.