bjornmelin/dev-skills
51 skills361 installs255 starsGitHub
Install
npx skills add https://github.com/bjornmelin/dev-skillsSkills in this repo
1Grill MeThis skill interviews the user about a plan or design until reaching shared understanding. Developers use it to stress-test a design by resolving each branch of the decision tree one question at a time. It recommends an answer per question and scores viable options with a weighted decision framework, exploring the codebase or web when a question can be answered that way.39installs2StreamdownStreamdown is guidance for Vercel's Streamdown library, a react-markdown replacement built for streaming AI output. It shows how to render AI-generated markdown in React chat UIs while gracefully handling incomplete syntax during streaming. A developer uses it when building chat interfaces with the AI SDK useChat/streamText, or migrating from react-markdown. It covers Shiki code themes, KaTeX math, Mermaid diagrams, styling with Tailwind, and hardening output with rehype-harden.17installs3Vitest Devvitest-dev guides writing high-signal, low-flake, fast Vitest test suites for TypeScript and Next.js. It follows a defined procedure: map the unit under test, pick the lightest test, design a minimal matrix, implement, harden against flakiness, and optimize. A developer uses it when adding or improving tests and shaping Vitest config for local developer experience and CI throughput. It covers pool choice, isolation, parallelism, caching, and Next.js 16 integration defaults.13installs4Zod V4Zod v4 provides guidance for schema validation in TypeScript using Zod v4. It covers designing schemas, migrating from Zod 3, handling validation errors, and generating JSON Schema or OpenAPI. A developer uses it when defining data contracts or integrating validation with React Hook Form, tRPC, Hono, or Next.js. It documents v4 APIs such as top-level string formats, strictObject/looseObject, registries, branded types, and recursive schemas.13installs5Docker ArchitectThis skill designs, implements, and hardens Docker images and Compose environments. Developers use it to create or rewrite Dockerfiles and docker-compose files, audit existing container setups for security and size issues, and add CI pipelines for build, test, scan, and publish. It favors multi-stage BuildKit builds and least-privilege runtime defaults.12installs6Dmc PyThis skill provides guidance for building Dash applications with Dash Mantine Components v2.x. Developers use it when creating dashboards, forms, and data-visualization apps to pick the right component and wire up callbacks. It covers MantineProvider theming, style props, pattern-matching and clientside callbacks, multi-page apps, and charts across 100+ components.10installs7Pytest DevPytest-dev is a Claude Code skill for Python testing that writes and refactors pytest suites, fixes flakiness, designs fixtures and markers, and speeds up test collection and runtime. A developer uses it for pytest best practices, pytest 9.x features, plugin selection, and CI tuning such as GitHub Actions sharding and xdist parallelism. It ships reference guides and helper scripts for profiling slow tests and splitting suites by historical timings.10installs8Ai Sdk CoreAi-sdk-core is a Claude Code skill for building with AI SDK Core: generating text and structured output, tool calling, embeddings and reranking, MCP integration, middleware, telemetry, and error handling. It provides a function-selection table (generateText, streamText, generateObject, streamObject, embed, rerank) and patterns for typed and dynamic tools, multi-step execution, and provider setup. Developers use it to wire LLM calls, tools, and MCP servers with a consistent API across providers.9installs9Ai Sdk UiAi-sdk-ui is a Claude Code skill for building chat and generative UIs with the AI SDK React hooks. It provides a hook-selection guide (useChat for conversations, useObject for structured data, useCompletion for single-turn text) and patterns for status handling, error and retry, message editing, file attachments, custom request options, and message metadata. Developers use it to build streaming chatbots and tool UIs wired to Next.js, Node, Fastify, or Nest backends.9installs10Notebook Ml ArchitectThis skill audits, refactors, and designs machine-learning Jupyter notebooks with production-quality patterns. Developers use it to detect anti-patterns, data leakage, and reproducibility issues, then refactor notebooks into modular Python pipelines. It also generates ML workflow templates, adds seeding and environment capture, and converts notebooks to scripts.9installs11Codex Utilscodex-utils explains how to use Codex's built-in coordination, retrieval, and utility tools effectively during live sessions. A developer relies on it to maintain a step plan, ask short multiple-choice questions before risky edits, inspect local images, search or browse the web for source-backed information, discover MCP tools, and run independent tool calls in parallel. It prescribes using the lightest tool that resolves the need.8installs12Supabase TsSupabase TS gives production integration patterns for using Supabase in Next.js, React, and TypeScript apps. It shows SSR-correct server, browser, and middleware clients, plus auth, Row Level Security, storage, realtime, and Edge Functions patterns. A developer uses it when wiring Supabase into a Next.js App Router project and wants to avoid common auth and RLS mistakes. It also covers pgvector semantic search, Vercel connection pooling, and Zod v4 response validation.8installs13Aws ArchitectureAws-architecture is a Claude Code skill for planning AWS infrastructure and application integrations with production-grade service selection, boundaries, and rollout thinking. Its workflow reads the repo AGENTS.md, optionally runs a repo-inventory detect, consults AWS references, and prefers managed services and the simplest correct topology. It outputs an architecture recommendation, a service map, and a rollout and verification plan. Developers use it for AWS service selection and infra design, not repo-specific policy.7installs14Dash Auditdash-audit audits Plotly Dash apps for callback hazards, state-flow risks, layout and accessibility issues, and Dash-specific UX regressions. A developer uses it for a Dash callback review, Dash UI audit, or a read-first remediation plan. It runs a callback-map preflight, reads only the needed Dash references, and reports grouped findings ordered by risk. It is not for generic React or Next.js UI review.7installs15Deep Researcherdeep-researcher runs deep, cited research across official docs, Codex web tools, the Context7 API, GitHub, package source, rendered web pages, and Firecrawl, backed by evidence ledgers. A developer uses it for defensible, current, multi-source research such as library or API decisions, dependency investigations, release analysis, and high-stakes citations. It ties every claim to source IDs, freshness, and a confidence score, and stops with UNVERIFIED when sources disagree.7installs16Dmc Best PracticesThis skill gives best-practice rules for building Dash applications with Dash Mantine Components. Developers use it when writing, reviewing, or refactoring Dash apps to catch anti-patterns in callbacks, styling, theming, and performance. Rules are grouped by impact level and can be emitted as machine-readable UI audit findings.7installs17Gh Pr Review FixThis skill resolves GitHub pull-request review comments end-to-end. Developers use it to fetch unresolved review threads, apply minimal verified fixes file-by-file, run repo-native verification, and re-check until the PR is clean or blocked. It commits one scoped conventional commit and pushes when checks pass.7installs18Taste MetaskillTaste metaskill is a router that points frontend UI work to one focused design-taste reference so output is less generic and more distinct. It picks a single primary reference from options like baseline product UI, high-end marketing, minimalist, GSAP scroll, or brutalist dashboards. A developer uses it for web UI design, redesign, art direction, motion, layout, typography, and color work. It explicitly excludes backend and non-UI work and enforces anti-AI-slop guardrails.7installs19Technical WritingTechnical Writing guides the creation of clear internal engineering documentation. It covers specs, architecture documents, runbooks, API references, and changelogs, with copy-paste Markdown templates for each. A developer uses it when they need repository-grounded, actionable docs rather than marketing copy. It also adapts tone and depth to the target audience, from developers to stakeholders.7installs20Agents Md MaintainerAgents-md-maintainer is a Claude Code skill that decides when a repo's AGENTS.md should be updated after implementation work. It treats AGENTS.md as durable repo guidance, not a task log, and only allows edits when the code, tests, and docs already prove a new rule. It lists what qualifies (build/test commands, architecture invariants, canonical shapes, recurring failure modes) and what does not (one-off features, task plans, session narration). Developers use it as an end-of-implementation maintenance pass.6installs21Commitcommit stages and commits changes in semantic, reviewable groups. A developer uses it to organize commits or clean up a branch before pushing. It scopes to files from the current task, keeps each staged set to one semantic change, writes short conventional commit messages, and asks before staging unrelated or ambiguous changes.6installs22Docs AlignThis skill aligns documentation with code by detecting drift after implementation work. Developers use it to update ADRs, specs, README, and AGENTS files based on what changed in the code. It runs a drift-collection script, compares against docs, and produces a gap map of doc tasks rather than editing blindly.6installs23Langgraph MultiagentThis skill builds, audits, and migrates multi-agent systems using LangGraph and LangChain. Developers use it to design supervisor and orchestrator-worker topologies, wire tools, memory, guardrails, and MCP integration, and add observability. It stays version-accurate by resolving current APIs from docs and installed versions and helps migrate off deprecated patterns.6installs24Repo Context BuilderRepo-context-builder is a Claude Code skill that creates or refreshes REPO_CONTEXT.md and REVIEW_BRIEF.md documents for a repository. It inspects manifests, entrypoints, tests, CI, and architecture signals, then writes two high-signal documents so future agents and chats can work without re-discovering the repo. A developer uses it on a new repo, to refresh stale docs, or to produce a grounded task-scoped brief.6installs25Rust ExpertRust-expert is a Claude Code skill that acts as the default router for core Rust engineering, including implementation, debugging, code review, refactors, compiler errors, ownership and lifetimes, traits and generics, async and concurrency, typed errors, and Cargo and toolchain work. It inspects the repo first, loads only the references a task needs, and uses compiler and test feedback as design evidence. A developer uses it for general Rust work and it routes framework-specific tasks to narrower Rust specialists.6installs26Rust Tui RatatuiRust-tui-ratatui is a Claude Code skill for building Rust terminal user interfaces with Ratatui. It covers widgets and layouts, crossterm or termion event loops, async input, app state machines, keybindings, terminal UX, and snapshot tests. A developer uses it when building interactive TUIs that need persistent terminal state, keeping rendering pure and treating terminal cleanup as reliability-critical.6installs27Ai Sdk AgentsAi-sdk-agents is a Claude Code skill giving expert guidance for building autonomous agents with the AI SDK v6+ ToolLoopAgent. It covers when to use ToolLoopAgent versus core generateText/streamText, loop control with stopWhen and hasToolCall, structured output, streaming, and type-safe client integration. It also documents runtime configuration via callOptionsSchema, prepareCall, and prepareStep, plus multi-agent workflow patterns. Developers use it when creating agents, tool loops, or agent workflows with the Vercel AI SDK.5installs28AutoreviewAutoreview is a Claude Code skill that runs a Codex-only structured code-review closeout over local, branch, or committed diffs via a bundled helper. It uses codex exec for deterministic pass/fail JSON output and treats findings as advisory, requiring each to be verified against real code before changes. It supports target modes and flags for model, reasoning effort, web search, parallel tests, and output files. Developers use it as a final review gate before committing or shipping.5installs29Browser Workbench Setupbrowser-workbench-setup configures a new repository for browser-based UI work using playwright-interactive as the primary interactive tool and agent-browser for CLI smoke checks. A developer runs it when bootstrapping browser testing, screenshots, auth-state reuse, or local UI debugging in a fresh repo. It installs a repo-local Playwright dependency, sets up artifact directories and gitignore rules, and detects the auth provider before making changes.5installs30Bun Auditbun-audit is a router skill over a shared Bun audit/remediation engine. A developer uses it to audit a repo for Bun-first correctness, explain audit findings, list rules, plan safe fixes, apply low-risk remediations, or validate Bun-related changes. It delegates all rule and remediation logic to the bun-dev shared platform and creates rollback artifacts for safe fixes.5installs31Bun Devbun-dev is a rule-first Bun guidance skill with a shared audit and remediation engine. A developer uses it when adopting Bun, migrating a repo from Node.js, fixing package management, optimizing scripts and monorepos, configuring Bun with TypeScript, or deploying Bun workloads on Vercel. It routes decisions through a prioritized rule set and exposes audit, plan-fixes, apply-safe-fixes, and validate commands.5installs32Caveman Compresscaveman-compress rewrites documentation and prose files such as AGENTS.md, todos, preferences, and repo notes into terse caveman-speak to reduce input tokens. A developer uses it to shrink dense markdown while preserving code blocks, URLs, file paths, and heading structure. It works directly in the active session, auto-applies only high-confidence matches, and leaves source code files untouched.5installs33Codex Sdkcodex-sdk gives architect-level guidance, workflows, and scripts for building agentic coding systems with OpenAI Codex. A developer uses it for the Codex SDK (@openai/codex-sdk) threads and streaming, Codex CLI automation (codex exec), MCP server usage, multi-agent orchestration via the OpenAI Agents SDK, and durable state in SQLite. It emphasizes structured outputs, audit trails, and safe-by-default sandbox and approval patterns.5installs34Context7 Researchcontext7-research performs deep library, dependency, framework, and API documentation research using the Context7 MCP tools. A developer uses it to get up-to-date docs, version-specific API reference, migration guidance, or to verify code against official documentation. It resolves a library ID, queries docs within a strict 3-call budget, and flags gaps as UNVERIFIED.5installs35Convex Auditconvex-audit performs read-first audits of a Convex-backed codebase, covering schema quality, security, runtime boundaries, migrations, and function-surface risks. A developer uses it for a Convex review, backend audit, or remediation plan on existing architecture. It runs convex-scan inventory, surface, and gaps, then outputs ranked findings and an implementation-ready remediation checklist.5installs36Convex Feature Specconvex-feature-spec generates a complete Convex-first feature specification for a repo, including data model, API surface, rollout, and verification. A developer uses it for green-field or breaking-change-tolerant feature planning where the output is a full implementation spec. It grounds the design in the current backend via convex-scan inventory and outputs schema and function contract changes with a rollout plan.5installs37New BranchThis skill creates and switches to a new conventional-commit and semver-compliant git branch before any work begins. Developers use it when they want a scoped branch prepared first and a clear plan laid out in chat. It stops after switching branches unless asked to implement, then presents the intended changes, key files, verification, commit shape, and risks.5installs38OpensrcOpensrc is a Claude Code skill that inspects the actual source code of a dependency using the opensrc CLI instead of relying only on docs and types. A developer invokes it when behavior, migration risk, or implementation internals matter, using commands like opensrc path and git diff between pinned versions. It supports npm, PyPI, crates.io, and Git repo sources and includes a dependency upgrade-audit workflow.5installs39Platform ArchitectPlatform-architect is a Claude Code skill that routes broad full-stack platform work across Next.js, Expo, Convex, and monorepo tooling. It runs a deterministic repo inventory to detect the stack, selects the right lane (web, native, or full-stack), then routes narrow tasks to downstream specialist skills. A developer uses it when a task needs stack detection, platform-level planning, or cross-cutting architecture rather than a single narrow change.5installs40Repo Docs AlignRepo-docs-align is a Claude Code skill that syncs repository docs such as AGENTS.md, README, ADRs, specs, runbooks, and doc comments with the current code and workflow. It anchors on the current worktree, inventories doc surfaces, and turns drift into a grounded plan or a verified doc-alignment implementation. A developer uses it after a big change, when docs have drifted, or when governance requires an execution plan.5installs41Rust Cli ClapRust-cli-clap is a Claude Code skill for building Rust command-line tools with the clap library. It covers clap derive and builder APIs, subcommands, flags, config and env precedence, stdout/stderr contracts, JSON output, exit codes, completions, and manpages. A developer uses it when working on scriptable, terminal-friendly Rust CLIs and their tests and binary distribution.5installs42Rust Mega EngRust-mega-eng is a Claude Code skill that acts as an explicit-only orchestration layer for broad Rust architecture and product engineering. It maps the product surface, identifies crate and service boundaries, chooses mature crates and frameworks first, and splits delivery into reviewable lanes with a decision ledger. A developer invokes it explicitly for multi-crate workspace strategy, large refactors, release engineering, and end-to-end Rust product planning across CLI, TUI, Tauri, services, and libraries.5installs43Rust Tauri AppsRust-tauri-apps is a Claude Code skill for building Tauri v2 desktop and mobile applications with a Rust backend. It focuses on small Rust command surfaces, secure least-privilege capabilities and permissions, typed IPC contracts, app state, plugins, and distribution. A developer uses it when working on src-tauri commands, capabilities, secure IPC, sidecars, updater, and bundling for a Tauri app.5installs44Rust Web ServicesRust-web-services is a Claude Code skill for building Rust HTTP services with Axum, Tokio, Tower, and Hyper. It covers router setup, middleware layers, request extractors, typed errors, tracing, config, graceful shutdown, SQLx data access, and background jobs. A developer uses it when building production Rust REST APIs, keeping handlers thin and treating errors, tracing, timeouts, and shutdown as part of the feature.5installs45Ship BranchShip Branch finalizes a git branch by staging only the files related to the current task and splitting the work into separate conventional-commit commits. It then pushes the branch and opens a pull request into main with a conventional-commits title and reviewer-ready body. A developer uses it when they want to commit, push, and open a PR in one guided flow. It can also scaffold missing repo PR and release-default files when asked.5installs46Streamlit Master ArchitectStreamlit Master Architect provides architect-level guidance for building, refactoring, debugging, testing, and deploying Streamlit apps. It enforces correct rerun, session state, caching, and fragment patterns for single-page and multipage apps. A developer uses it when building production Streamlit data apps and wants AppTest and Playwright MCP testing plus security-by-default. It ships templates, scaffolding scripts, and an evergreen mode that audits the installed version rather than guessing APIs.5installs47Upgrade Pack GeneratorUpgrade Pack Generator creates repo-local dependency-upgrade packs before any code changes are made. Each pack contains a canonical manifest plus a playbook, trigger prompt, and operator mode rendered from it. A developer uses it when an agent needs to research an anchor package, expand related dependencies, detect repo context, and produce a reusable upgrade folder. It is monorepo-aware and produces docs and research only, deferring implementation to a later run.5installs48Gh Deps IntelThis skill plans and executes dependency upgrades for JavaScript/TypeScript and Python repos, including monorepos. Developers use it to audit outdated packages, mine GitHub releases and changelogs for deprecations and breaking changes, and produce a Markdown plus JSON upgrade report. In execute mode it applies the planned version bumps and runs repo-native verification.3installs49Opensrc InspectOpensrc-inspect is a Claude Code skill for fetching, inspecting, and managing local opensrc source snapshots of packages and GitHub repositories. A developer uses it when API docs alone are not enough and source-level evidence is needed for a migration, bug, or compatibility question. It defaults to non-mutating --modify=false fetches and treats the opensrc/ directory as read-only source context.3installs50Native Accessibility PerformanceThis skill handles motion accessibility and UI-thread performance for Expo and React Native apps. Developers use it to add reduced-motion and haptics behavior, audit dropped frames and frame pressure, and validate gestures on device. It includes an audit CLI and a manual device-proof checklist, and defers implementation patterns to adjacent native-motion skills.1installs51Native LottieThis skill integrates Lottie animations into Expo and React Native apps. Developers use it for lottie-react-native and dotLottie assets, asset bundling, playback refs, and unmount cleanup. It also covers accessibility labels, reduced-motion behavior, and iOS/Android validation, with an audit CLI to surface issues.1installs