Recently updated open-source projects

color-combo-skill helps an agent choose a palette by product mood rather than picking arbitrary swatches. Its catalog covers premium, futuristic, organic, minimalist, gothic, tropical, gradient, and other directions, with names and usage guidance for each combination. It works as the color stage in a modular design chain and can hand off next steps to typography, effects, and icon skills.

An interactive setup adds test infrastructure to TypeScript or Python projects, while one CLI dispatches Playwright web tests, Appium mobile tests, or k6 load tests. AI planner, generator, and healer definitions are currently focused on TypeScript, and several other test types, including API, accessibility, and visual regression, are still planned.

Agent Memory System creates repository-local Markdown and JSON artifacts covering project structure, commands, interfaces, security context, dependency graphs, worklogs, and handoffs. Coding agents can read the generated index and query the graph before editing shared code, then record checkpoints for the next session. It is not a chatbot memory service or agent runtime, and its benchmark results are maintainer-run early measurements rather than independent validation.

agent-terminal lets an AI agent launch interactive CLI applications, capture their screen as ASCII text, and send keyboard input without a display. It can be used as an npm library or exposed through an MCP server for CLI testing, monitoring, and build automation. Built-in safeguards include command allowlists, workspace path validation, environment sanitization, input checks, and session limits.

claudebox launches a Claude Code project through bubblewrap on Linux or sandbox-exec on macOS. It shadows most of the home directory, mounts the project parent read-only, and blocks sensitive runtime sockets by default, with opt-in SSH, GPG, and XDG access. The project explicitly says this is not a complete security boundary; Linux is stable and macOS experimental.

Subtask adds a skill and CLI to Claude Code for creating tasks and running Claude, Codex, or OpenCode subagents in parallel, with a separate Git worktree for each task. Its terminal interface shows status, diffs, and conversations, while the lead agent reviews completed work and asks before merging. The project is still in early development, so installation and behavior may continue to change.

The plugin turns an idea into numbered acceptance criteria, decomposes the work into a dependency-ordered task graph, and runs tasks in parallel isolated Git worktrees with test-driven implementation. Code review and requirement-level verification gate atomic merges, while disk-backed state supports recovery after crashes or context resets. Remote pushes still require explicit user approval.

Wreckit takes a text backlog and moves each item through codebase research, planning, story-by-story implementation, and pull request creation for you to review. State and artifacts stay in Git-trackable JSON and Markdown files, making runs inspectable and resumable. It requires an AI agent or API access and can consume substantial tokens; an optional Sprite-based sandbox isolates riskier execution.

YouTube Transcript Ingestion filters a channel by title keywords and date range, then saves one readable Markdown transcript per matching video. It confirms the list before fetching, supports an optional notes folder for finding genuinely new material, and offers pacing or browser-cookie strategies when YouTube throttles requests. The output includes title, date, URL, source, and transcript fields, making it easy to search in Obsidian or feed into another tool. yt-dlp and MarkItDown’s YouTube extra are required.

blpapi-expert helps Claude write Bloomberg Desktop API code for reference, bulk, historical, intraday, subscription, derivatives, override, and error-handling workflows. It consults bundled request schemas and canonical examples instead of guessing field shapes, while keeping xbbg version differences visible. The scope is local Desktop API sessions attached to a Bloomberg Terminal; B-PIPE, Server API, Platform, EMSX, and Excel COM are excluded. Running the generated code requires a logged-in Terminal plus Bloomberg's SDK and xbbg.

Copy URL registers itself as a macOS browser, but selecting it copies the incoming address to the clipboard and immediately quits instead of loading the page. This is especially useful with browser picker apps such as Velja, where it appears alongside installed browsers and turns one selection into a copied link. The app requires macOS 13 or later and is installed by downloading the release, unzipping it, and moving it into the Applications folder.

Install this Claude Code skill to work through schematic choices, layout, power, RF, high-speed signals, thermal design, EMC, DFM, compliance, testing, and production checks in natural language. It supports planning a new board, reviewing a design before fabrication, and troubleshooting hardware issues, with workflows for LCEDA Pro and KiCad. The skill supplies design guidance; direct EDA control requires a separate automation skill.

Turn an enterprise architecture request into a decision rather than a flat list of technologies. The skill begins with focused discovery around business drivers, quality attributes, constraints, scale, and audience, then commits to one recommendation while naming its tradeoffs and rejected alternatives. It can structure ADRs, capability maps, target architectures, roadmaps, standards, and audience-tiered decks, but it is an instruction skill rather than an MCP server or external integration.

jobspipe-cli exposes the JobsPipe API through a standalone CLI and agent skill. One command searches normalized live job postings from more than 30 sources with filters for title, location, seniority, company, and recency; the other detects the technology stack behind a domain. Both print JSON, and usage requires a JOBSPIPE_API_KEY, with a stated free quota of 5,000 requests per month.

embedded-uiux-walkthrough guides an agent to add tutorial routes or embedded tours to a web app using real components rather than screenshots. It covers demo-only data, no-write flows, fake cursors, target highlights, click and typing animations, iframe scroll or input locking, and mobile-aware targeting through stable data-tutorial hooks. Use it for onboarding, customer demos, or in-app training.

Have the agent emit semantic JSON instead of raw HTML, then render it as a self-contained file with styles, scripts, and fonts included. Reports, diagrams, or documents can be stacked into a persistent dashboard and restyled later without changing their content. The files open offline and can also be published through GitHub Pages or a custom deployment.

agent-runbook lets skill authors describe multi-agent workflows in YAML and compile them into SKILL.md files with checkpoint support. Steps declare typed inputs and JSON Schema outputs, while the compiler checks references, dependency closure, and cycles before producing the artifact. Runbooks can dispatch agents, run scripts, branch, loop, execute parallel work, and resume from a checkpoint; the project notes that its API is still changing.

meta-skills is a collection of 17 standalone process skills for Claude, including rubber-duck questioning, steelmanning, pre-mortems, first-principles analysis, red-teaming, scope cutting, context budgeting, distillation, handoffs, and teach-back. Each skill is designed to challenge assumptions or verify work before trust. Install the collection globally, per project, or copy a single folder when you need only one method.