Recently updated open-source projects

Fashion Item Extractor breaks an outfit, runway, street-style, or lookbook photo into identifiable garments and accessories, then produces a clean product image for each item. Its default workflow writes one prompt per item, crops a source rectangle as a visual anchor, and sends the crop plus prompt to Gemini or Seedream for image-to-image generation. Users can stop after analysis or prompts; final rendering needs the chosen provider key and optional Python packages.

webEmbedding starts with a URL, checks whether the page can be reused directly, and otherwise captures browser state before generating a bounded HTML, CSS, React, or Next.js scaffold. It compares the original and candidate through screenshots, DOM, computed styles, interaction states, and responsive breakpoints. The tool is aimed at static or semi-static pages; login-only surfaces, captchas, payments, and server-side app behavior remain out of scope.

This Claude Code skill detects Russian bureaucratic phrasing, English calques, inflated claims, chatbot artifacts, structural habits, and typography patterns. It offers modes for social posts, business letters, personal messages, copywriting, and AI-style audits, with optional voice calibration from sample texts.

Darce is an account-based terminal coding agent that can read, write, and edit files, search a codebase, run shell commands, and fetch web content after installation and login. It supports in-session model switching or automatic routing, plus context compaction, resumable sessions, and cost tracking. The free plan limits monthly requests and exposes only a subset of read-oriented tools.

Technical-Explanatory Minimalism guides agents in making illustration-led technical diagrams, schematics, cutaways, flow figures, and interactive explanations. Interactive examples are self-contained HTML files with inline SVG and vanilla JavaScript, so they can be opened without a build step and themed with CSS variables. The rules emphasize showing mechanisms, coupling views and controls, annotating in place, and keeping data honest. Verified examples cover engines, state machines, exploded views, gears, and linked plots in light and dark modes.

Judgment over Velocity helps product teams decide whether an idea deserves building, buying, or further testing. It applies a spreadsheet test, an 80/20 structural cut, figure-ground analysis, and evidence probes that rank real demand signals above stated preferences. Each gate may kill the idea and records an epitaph; survivors receive the cheapest disconfirming test and an explicit kill condition. The skill is designed for CPO-level choices and treats “build nothing” as a legitimate verdict.

Bellwether creates a readable daily brief from congressional STOCK Act disclosures and institutional SEC 13F filings, then adds a watchlist and a paper-trading Ledger. Users can compare copied trades with the S&P 500 through backtests, inspect momentum signals, and ask Claude to fill analyst targets or ratings for watched stocks. It uses free public data and the Ledger does not place real trades. The project is educational and explicitly not financial advice.

visual-review performs visual QA on rendered PDFs, slide decks, reports, and HTML pages. It looks for clipping, overflow, overlaps, broken images, typography and contrast problems, density imbalance, and page-count mismatches, then labels findings as HOLD, FIX IF TIME, or NOTE. A Quick Pass handles short documents; Full Review needs editable source and shell access, while unavailable renderers trigger a text-only fallback.

site-visibility-report measures how a website appears in classic search, AI citations, and direct-answer surfaces, then renders the findings as a client-facing HTML and PDF report. You choose the language, measurement depth, and page scope before running its collection and scoring workflow. Bundled templates, a standalone-file builder, and headless Chrome rendering handle delivery; the manual path needs the listed Python packages.

ClaudeGuard is a report-only pre-merge policy gate. It reads a branch diff, selects matching core and project rules, evaluates added lines, and emits PASS, WARN, or FAIL plus parseable JSON. Version 0.2.0 adds reproducible candidate detection and exemptions for selected rules, fixture tests without an API key, and a runner verdict line. The gate never edits, stages, or commits code; CI maps FAIL to a blocking exit.

AI Humanizer detects which English sentences appear machine-written, rewrites only those passages, and verifies protected terms, numbers, citations, and domain names after each round. It uses a keyless detector as a proxy rather than a separate paraphrase API, and documents that scores vary by domain and tool. The project frames it for legitimate editing and research, subject to institutional and publisher policies.

The skill first asks about the audience, occasion, source material, and available numbers, then combines methods from nine reporting families into an outline for annual reviews, project retrospectives, promotion cases, or resource requests. It produces a presentation-ready HTML report, with optional editorial or Swiss-style web decks when guizang-ppt-skill is installed; missing figures are marked for completion rather than invented.

Ultimate Meta Ads Skill turns Claude Code into a structured Facebook and Instagram advertising strategist. It combines evergreen strategy, copy, creative, funnel, scaling, and troubleshooting frameworks with a dated supplement that updates tracking, objectives, budgets, attribution, learning phases, and creative specifications for recent Meta behavior. Checklists and templates support launch and optimization. Because ad mechanics change quickly, the project recommends verifying current details against Meta’s official documentation and the advertiser’s own data; results are not guaranteed.

Myra Agents skills is a small catalog of installable skills for common agent workflows. It includes a searchable code index, runtime debugging with direct logs, launch-content drafting from changelogs, persistent-memory consolidation, and accessibility-tree tests for native macOS app surfaces. Each skill is a self-contained folder that works with compatible agents such as Claude Code, OpenCode, Codex, and Cursor, and can be installed individually or as a set.

G Desktop Suite turns Google Drive into a focused desktop app for Windows, macOS, and Linux, separating your files and documents from a crowded browser session. It follows the operating system's dark-mode setting and can be installed from release packages or supported package managers. The original author stopped maintaining the project in 2021 and invited new maintainers, so it is better suited to users who accept possible compatibility and update gaps.

Enter workload details through Claude Code commands to receive a 7R recommendation, blocker triage, readiness scoring, and focused assessments for containers, databases, networks, or VMware estates. The skill can also draft wave plans, cutover runbooks, portfolio reports, and rough cost estimates. Its methodology and guardrails are stored locally, so no MCP server is required.

GitHub Boards lets an agent manage a GitHub Projects v2 board by conversation: file issues, inspect human and agent queues, move lanes, route owners, sync TODOs, reconcile drift, and plan undo from snapshots. Every write is staged for explicit approval, and the tool fails closed on missing or ambiguous configuration. It requires authenticated gh, Node 18+, and project permissions.

lacuna-toolkit exposes the Lacuna Music API through a TypeScript SDK, a command-line tool, an MCP server, and an Agent Skill. Developers can generate music from Node 18+, shell scripts, an MCP-compatible client, or an AI coding agent, while the SDK also handles webhook verification. Using any surface still requires access to the Lacuna service and its credentials.