Recently updated open-source projects

Founders raising from pre-seed through Series B can use this Claude skill to draft cold emails, follow-ups, investor replies, meeting recaps, and due-diligence preparation. It asks for the company and round details before writing, leaves visible placeholders when facts are missing, and refuses to invent investor research or fake competitive pressure.

cc-provider-add configures MiMo, GLM, Kimi, Bedrock, or another Anthropic-compatible endpoint for Claude Code using a separate `CLAUDE_CONFIG_DIR` per provider. It creates protected secret files, model settings, onboarding metadata, a launcher, and a shell alias, so multiple providers can run in separate terminals without changing `~/.claude/settings.json`. The skill verifies the connection with a real API ping. It is designed for terminal users of Claude Code, and trusting a third-party endpoint means sending prompts to that provider.

win11-admin gives Claude Code PowerShell procedures for Windows 11 Pro and Enterprise administration, including registry changes, services, Group Policy, Defender, firewall rules, local users, updates, debloating, and performance tuning. Its guidance emphasizes creating a restore point, backing up affected registry keys, supplying rollback commands, and identifying components that should never be disabled before making system-level changes.

best-aeo-skill audits a URL with 33 local evidence collectors, scores technical accessibility, content citability, structured data, and entity signals, and reports how the site may be cited by ChatGPT, Claude, Perplexity, Gemini, and other answer engines. Its fix commands can draft `llms.txt`, robots rules, JSON-LD, and content changes as a reviewable diff, while monitor mode tracks score regressions. The tool runs without an external API and does not push changes without confirmation.

The skill selects region-appropriate Jito endpoints, recommends tips from live percentile data, and submits signed transactions or bundles with polling. It also analyzes bundle siblings, KOL tipping patterns, and possible sandwich attacks; Jito and Helius Sender are covered, while several alternative relay integrations remain stubs, and analysis requires a Solana RPC.

Antiflow acts as a constrained AI tech lead: it creates the project structure, configuration, and routine setup, but leaves core features as clearly marked user-owned TODOs collected in a task file. Assistance is calibrated by role, experience level, and a chosen AI contribution ratio. It can persist settings across sessions, switch to a Socratic student mode, generate tests for the withheld logic, and report how much domain code came from the developer versus the agent.

docker-wsl-cleanup reports the Docker virtual disk and `docker system df`, guides a confirmed prune, stops Docker Desktop cleanly, shuts down WSL, and prints the appropriate VHDX compaction command for Windows Pro, Enterprise, or Home. It restarts Docker and reports the reclaimed space after the user completes the administrative step. The skill requires Windows 10 or 11 with Docker Desktop using WSL2 and an administrator PowerShell; it never self-elevates or force-kills Docker.

lets.skill distills public posts, interviews, and livestream material about the Chinese internet phenomenon known as Wanxue into response and expression frameworks. It offers role-play, researcher-oriented event analysis, trap-question detection, and a separate language-literature branch. The project states that it is not authorized or endorsed by the person represented, uses material current through 2025, and treats undisclosed topics as framework-based inference rather than a real private reaction.

The skill converts Markdown through Pandoc’s Typst writer, combines the result with a selected theme, and compiles a printable PDF. It ships with two Claude-inspired themes and supports custom themes described in DESIGN.md; Pandoc 3.2 or newer and Typst must be installed on the system.

wechat-article-down opens a WeChat Official Account URL through chrome-devtools MCP, waits for the article to render, scrolls through lazy-loaded media, and extracts the body plus metadata. It converts the HTML to Markdown and saves images, audio, and optionally video alongside the article, with batch processing supported. Node.js 16 or newer and browser automation are required, while video downloads additionally use yt-dlp.

The Douglas Adamiser is a Claude Code writing skill calibrated against roughly 528,000 words from ten major Douglas Adams books. It applies measured patterns in diction, sentence length, pacing, and bathos rather than recycling quotations or familiar imitation clichés. Deterministic validation scripts help check whether a rewrite follows the documented style constraints.

apca-contrast-skill bundles a deterministic Python implementation of APCA-W3 that measures Lc for text and background pairs, looks up passing font sizes and weights, audits Tailwind v4 light and dark tokens, and reverse-solves a color for a target Lc. It accepts common CSS color notations and needs no third-party Python package. APCA is still a WCAG 3 draft model, so legal accessibility work also needs a WCAG 2.x ratio check; alpha, wide-gamut displays, typography, and vision conditions need additional judgment.

jackson-wang.skill distills public interviews, posts, performances, and works into five mental models, eight decision heuristics, and a style reference for advisory or role-play conversations. It can switch among Cantonese, English, Mandarin, and Korean and recognize recurring references. The project explicitly treats its answers as inferences from public material rather than Jackson Wang's real opinions; its research cutoff is April 2026 and private thoughts are out of scope.

Council selects three thinking personas from a pool of five based on the user's situation, gathers their independent views, and lets a JUDGE synthesize them while representing the two absent perspectives. It offers a lightweight role-play mode and a heavier mode with three parallel executions, using Regret, Reversibility, and Alignment as evaluation axes. The skill supports personal choices such as career, relationships, money, and ethics; it is not for fact lookup, technical implementation decisions, decisions for someone else, or urgent life-and-death situations.

Apply a repeatable engineering cycle to coding tasks regardless of language or technology stack. The skill emphasizes defining the specification before implementation, test-driven development, incremental delivery, and explicit verification before shipping. It also reviews changes across correctness, readability, architecture, security, performance, reliability, and operability to catch scope creep and recurring failure patterns.

humanizer rewrites text to sound more natural without changing API names, product names, or version numbers. It detects 18 patterns such as inflated vocabulary, filler transitions, passive voice, excessive em dashes, vague attribution, and formulaic conclusions, then adjusts the tone for design documents, READMEs, code comments, emails, or Slack. It works as a Claude Code or opencode skill and focuses on language style rather than technical correctness.

Urban Experience UI guides an agent from a chosen city and vibe to a complete frontend brief. It researches architecture, lighting, street rhythm, and materials, records design decisions, builds a storyboard and compiled specification, then implements HTML, CSS, and JavaScript with responsive and reduced-motion considerations. The city is treated as research input rather than a visual template to copy, and the workflow includes checks intended to prevent repetitive or low-quality generated layouts.

Blueprint converts a large coding objective into a Markdown plan whose steps can be executed by fresh agents in separate sessions. Every step carries its own context, tasks, branch or worktree guidance, validation commands, and rollback instructions. The plan also records dependencies, parallel opportunities, progress changes, and an adversarial review, making multi-session refactors or migrations easier to resume without relying on conversation history.