Recently updated open-source projects

openskill reads signals such as CLAUDE.md, package.json, and framework configuration to understand a project before looking for relevant Claude skills. It searches installed sources, community repositories, and an official registry when available, filters duplicate suggestions, and lets the user choose global or project-scoped installation. Searches can also be narrowed to areas such as frontend, backend, testing, or API design.

This repository currently centers on API Design Reviewer, a skill for analyzing OpenAPI or Swagger specifications, GraphQL schemas, and gRPC proto files across ten design dimensions. It produces scored Markdown and interactive HTML reports that explain what should change and why. Evaluation cases are included to check the consistency of the review behavior.

overnight-skill audits a codebase, asks the user to confirm protected files, and then launches fresh Claude Code sessions from a tmux runner to fix queued issues. Each iteration runs verification, commits and pushes to a dedicated branch, and reverts on failure, while a draft PR, progress log, and morning review capture the outcome. It runs headlessly with bypassPermissions, so its shell-enforced safeguards and scope require careful review before use.

inbox-triage scans an inbox directory, classifies each file as a project, learning item, experiment, archive candidate, or related asset, and presents the plan before changing anything. After confirmation, it can scaffold missing project folders, prepend triage metadata, move the files, and verify the result. It expects a workspace with the documented inbox, project, learning, sandbox, archive, and optional template directories.

crossfire runs a four-phase actor-critic workflow in which Claude plans and reviews while Codex implements and audits. Task tiers scale from quick edits to architecture-level changes, with deterministic tests, linters, debate, and fix-and-reaudit loops added as risk increases. Both Claude Code and Codex CLI are required, and the documented setup includes Windows-specific defenses.

Flowise Workflow Generator turns a workflow description into an importable Flowise JSON file, step-by-step UI instructions, and a Mermaid or SVG diagram of node connections and data flow. It covers Chatflows plus AgentFlow V1 and V2, with templates for chatbots, RAG, API integrations, multi-agent supervision, database agents, and human approval. Users can adapt the generated structure to their chosen models, tools, and data sources.

PixelPitch converts a LinkedIn visual brief into a prompt package tailored to the selected image model. It can include a main prompt, negative prompt where supported, dimensions and settings, a brand-alignment check, two creative variants, and accompanying post copy. Marketers can also provide brand colors, style preferences, audience details, and visual exclusions for more consistent output.

Guide Claude Code through the Python package lifecycle, from a src-layout scaffold and dependency setup to API naming, structured errors, tests, documentation, deprecations, and PyPI releases. The conventions use tools such as uv, pytest, rich, and MkDocs. The skill can also inspect existing projects for common packaging mistakes and configure pre-commit checks.

Weaver Mind looks across domains such as biology, architecture, game design, jazz, and physics for patterns that may apply to a problem after the obvious playbook has run out. Its report explains why each analogy transfers, how to act on it, the tradeoffs, effort, and potential impact, then ends with a wildcard and the smallest promising experiment you can run within seven days.

This Claude Code skill inspects routes, request parameters, response types, middleware, and validation schemas to produce an `openapi.yaml` file without requiring decorators or annotations. It supports common JavaScript, Python, Go, Java, and Rust web frameworks, maps native types into OpenAPI schemas, and detects authentication patterns such as bearer tokens, API keys, and cookies. The generated specification can be validated with Swagger CLI or Redocly.

Describe the entities, relationships, and ownership rules in natural language, and this Claude Code skill generates SQL migrations with Row Level Security, Supabase-style TypeScript types, Zod schemas, and Next.js App Router endpoints. It can also scaffold authentication choices, React forms, and many-to-many junction tables. Built-in checks call out common mistakes such as missing delete policies, unprotected junction tables, exposing the service-role key, or using imprecise numeric types for currency.

This Claude Code skill can turn a plain-language matching requirement into a regular expression with an explanation and test cases. It can also break down an existing pattern or run it against supplied strings, showing match positions and capture groups. Output adapts to JavaScript, Python, Go, or Java syntax, includes ready-to-use match or replace snippets, and warns about issues such as catastrophic backtracking and greedy matching.

Env Doctor scans JavaScript, TypeScript, Python, Go, Dockerfiles, and Compose files for environment-variable references. It compares code with `.env` and `.env.example` to identify missing, undocumented, or unused values, then checks common secret patterns and validates formats such as URLs, ports, email addresses, booleans, and timeouts. When no example file exists, it can generate a categorized `.env.example` template for the project.

This Claude Code skill analyzes the project's language, framework, package manager, and service dependencies, then creates a multi-stage Dockerfile, production and development Compose files, `.dockerignore`, and an environment template. It separates hot-reload development needs from a smaller production image and applies practices such as non-root users, fixed image tags, health checks, and excluding secrets from the build context. The result can be checked with `docker compose config`.

This Claude Code skill scans migration files, raw SQL, or ORM schemas to extract tables, columns, constraints, indexes, and relationships. It supports Prisma, Drizzle, TypeORM, Supabase, Knex, Django, Alembic, Go migrate, and SQL DDL. The generated `docs/database.md` includes a GitHub-renderable Mermaid ERD, per-table field documentation, relationship summaries, and index listings. Incremental runs update changed tables while preserving descriptions added by users.

This Claude Code skill inspects the language, framework, package manager, and monorepo layout before generating CI steps for linting, type checks, tests, builds, and optional deployment. GitHub Actions is the primary fully supported platform, with template support for GitLab CI and CircleCI. It can also add Dependabot configuration, minimal permissions, pinned action versions, OIDC authentication, path filtering, and stack-specific caches, then validate the workflow with `actionlint`.

This Claude Code skill reads a configurable Git range and converts commits into readable release notes instead of copying log messages verbatim. It categorizes Conventional Commits, highlights breaking changes, detects versions from common project files or tags, and writes entries in Keep a Changelog format. It can create or append to `CHANGELOG.md`, generate commit links when a repository URL is known, and produce separate changelogs for monorepo packages.

Describe an AWS problem such as recent 500 errors, authentication failures, or crashing ECS tasks. The skill selects relevant logs, builds a CloudWatch query, saves the result, and reports likely causes; it can also show health, alarms, and trends. A setup command discovers log groups, ECS resources, alarms, and the active region. Claude Code and a credentialed AWS CLI are required.