Recently updated open-source projects

Use this Claude Code plugin to scaffold or extend Expo React Native apps on a fixed TypeScript architecture. Its agents cover app setup, feature work, standards review, simulator or end-to-end runs, and EAS build, update, and submission; feature work must pass review before and after implementation. The locked stack includes Redux Toolkit, RTK Query, React Navigation, react-hook-form with Yup, i18n, and Maestro, so it suits teams willing to adopt those choices.

dependabot-triage handles npm, pnpm, yarn, and bun alerts by ranking impact, exposure, and CVSS, mapping import sites, selecting a minimal patched version, and checking the changelog for breaking changes. It updates every package-manager override in play, regenerates lockfiles, and aborts when resolved versions drift across local and CI tooling. The skill pauses for explicit approval before applying a bump or opening a pull request.

Use this skill to route marketing tasks across Google Ads, Google Analytics 4, and Google Tag Manager, combining them for campaign analysis and changes, funnel diagnosis, tag debugging, conversion reconciliation, and weekly reviews. Each service still requires its own SDK, MCP connection, and authentication. GA4 access is read-only, campaign creation is recommended through the Ads web interface, and conversion totals across the three systems should be reconciled rather than forced to match.

SPRUCE SQL Advisor grounds Claude in AWS Cost and Usage Report 2.0 schemas and the green-operations columns added by SPRUCE. It covers nested arrays, BILLING_PERIOD partition semantics, Athena and DuckDB syntax differences, and the meaning and units of carbon, electricity, and water fields. The skill is aimed at FinOps and GreenOps queries against this specific data shape, rather than general-purpose SQL assistance.

gh-tendr turns private GitHub Issues into an asynchronous review queue where Claude polls comments, classifies intent, and can reply, edit documents, implement changes, or request clarification. It requires an authenticated gh CLI with write access to a private repository, and destructive actions such as merges, force pushes, or production changes still need explicit confirmation; the project is pre-1.0.

Use prebuilt CLI verbs to inspect HTML, DOM state, console output, network traffic, screenshots, and scripted interactions against a live deployment. Rehearse can retain per-host authentication and scope Cloudflare Access credentials to the protected origin, while emitting stable JSON for CI, cron, or agent runners. It is Chromium-only and targets diagnostics and repeatable health flows rather than load testing or general synthetic monitoring.

The skill brings 13 frameworks into Claude Code for decisions around sales compensation, competitive moats, fundraising, go-to-market strategy, hiring, pricing, and AI-agent deployment. It can activate automatically when those topics arise or be invoked directly. The material was synthesized programmatically from more than 200 interview episodes rather than presented as a searchable transcript archive.

Forkcast expands a software goal into decision points, branches each choice, simulates up to ten steps, scores outcomes against weighted objectives, and emits a Markdown tree with a Mermaid diagram. It grounds steps in project artifacts, stops a branch when cumulative confidence falls too low, and sends selected paths to an adversarial critic. The tree is human-editable, so users can override scores and choose what to execute.

This extension applies Chinese plain-description principles to CRPG and choose-your-own-adventure dialogue nodes. It expects a structured input containing known player information, the foreground goal, background seeds, and one of five information types, then checks preconditions and node-specific rules for narration, NPC speech, player narration, or options. Load the general baimiao skill first; this layer adds the game-specific contract and regression checks.

Use this language-style skill to guide AI writing toward Chinese plain description (白描): concrete, perceptible details carry the scene while explicit emotion, judgment, and causal explanation are held back. It defines requirements, mechanisms, boundaries, distinctions from nearby styles, and a seven-point self-check. The skill is application-independent, so it can support fiction, screenplays, RPG dialogue, journalism, or other prose.

MagicMarkets API is a Claude skill and reference set for the MagicMarkets peer-to-peer sports exchange. It explains REST order placement, WebSocket live-price subscriptions, request-idempotency, heartbeat deadman switches, cancel and reconcile flows, and the exchange’s lay-price calculation. An API key is required for live requests and should stay in an environment variable. It documents the exchange’s trading operations and request flow.

Mentor adds an active teaching mode to Claude Code: it asks questions, offers progressive hints, explains concepts, and can patrol project changes while leaving implementation to the learner. It supports Socratic, tutor, or mixed teaching, multiple languages and humor styles, plus persistent user and project settings. Use it for deliberate practice and accountability, not for having the assistant write functional code.

Waypoint adds a structured planning layer to Claude Code. Work is split into independent milestones and strictly ordered phases; completion checks track acceptance criteria, deferred stubs, and architecture decisions, and a phase cannot start until the previous one is complete. A milestone remains open when required verification is missing or tracked stubs remain in the codebase, making it useful for projects that span multiple sessions.

Harden Node.js and Python dependency installs with release-age cooldowns, exact-version pinning, lifecycle-script restrictions, lockfile regeneration, and trust-policy checks. The skill audits current settings, can write machine- and project-level guards, and supports a dry-run or non-interactive CLI mode. It can also generate a Socket Basics pull-request workflow, which requires a Socket API key.

Use this skill to build a runnable digital persona from guided interviews and local text analysis rather than a large public dataset. It focuses first on tone of voice—sentence rhythm, recurring phrases, punctuation, emoji habits—and distinguishes private thoughts from outward wording, while adapting interview depth and checking contradictions. It is designed for ordinary people, including yourself, friends, colleagues, or family.

Harness Plan keeps a long development campaign resumable across Claude Code sessions using compact state, one active feature contract, session summaries, deterministic transition scripts, and risk-gated verification. The same /harness-plan command starts or resumes work, while optional change units split larger features into reviewable stages and companion discipline skills add more structured testing and safeguards.

Mmmmealplan turns a YAML profile and recipe library into a weekly menu with an aggregated, categorized shopping list. It uses live Migros promotions, nutrition, and allergen data while flagging macro gaps or restriction conflicts instead of silently changing the plan. The skill depends on the unofficial Migros MCP and does not place grocery orders.

This skill gives AI coding agents a worktree-based workflow for parallel feature branches. It initializes projects, uses Portless named local URLs, isolates local state, and supports safe resume, browser review, test-data cleanup, and branch finishing. It can also check whether independent local databases are available before parallel stateful work.