Tools Bench.

Product launches and open-source repos with enough signal to earn a second look.

Last Brew Time: Aug 19, 2026, 10:57 AM PT

Insight

AI builders this run are skipping the model entirely and shipping the workflow around it

01.AGENT-SKILLS LIBRARIES FOR CODING CLIS

02.AGENT MEMORY AND CONTEXT INFRASTRUCTURE

03.AI-DRIVEN DEVELOPER TOOLING

Featured

GitHub274.1K

obra/superpowers

An agentic skills framework & software development methodology that works.

youtube
Claude Code + SUPERPOWERS = The End of Vibe Coding? (Full Tutorial)

Market Signal

Why It Has Market Pull

Superpowers is one of the most established agentic-skills methodologies for coding agents on GitHub, built by a recognized engineer and adopted widely enough to spawn multiple independent forks and integrations across other IDEs. Growth has been tracked and reported by third parties over many months, an official plugin marketplace listing exists, and the project offers optional commercial support - real signals of a durable open source project rather than a one-off viral spike.

  • 274,188 verified GitHub stars and 24,548 forks (about 9% fork ratio) as of August 19, 2026
  • Repository created October 9, 2025, roughly 10 months old, with commits still landing the same day as this check
  • 291 open issues and 191 open pull requests, indicating a large, actively contributing community rather than passive stargazing
  • Spawned at least two independent public forks extending it into agent-agnostic tooling for other IDEs
  • Offers commercial support and is distributed via the official Claude plugin marketplace

feedbacks

What People Are Saying

  • "a complete software development methodology for coding agents, built on top of a set of composable skills"Dev.to article

  • "turn Claude into a proper dev workflow instead of just a chatbot"Blog review

  • "a 'game changer' for serious coding"Blog review

  • "Most publicly available skills don't just fail to help - they actively hurt by adding tokens, adding latency, and injecting constraints that make output narrower"Blog review (critical take)

  • "skills built by people who actually know the domain and have iterated on evaluation can produce output that's measurably sharper and more consistent"Blog review

  • "progressive disclosure to allow other editors to load skills as needed"Dev.to article

GitHub223.4K

mattpocock/skills

Skills for Real Engineers. Straight from my .agents directory.

youtube
mattpocock/skills: A complete AI Coding workflow, end-to-end

Market Signal

Why It Has Market Pull

This is a legitimately viral, actively maintained project from a well-known engineering educator, not a farmed repository. Independent snapshots taken over time show continuous organic growth alongside genuine press coverage, an official plugin marketplace listing, and real community discussion of specific skills - one of the most widely adopted agent-skill libraries in the current AI coding ecosystem and worth human inspection.

  • 223,558 verified GitHub stars and 19,236 forks (8.6% fork ratio) as of August 19, 2026
  • Repository created February 3, 2026 and still receiving commits the same day as this check - about 6.5 months of sustained, accelerating growth
  • 370 open issues, indicating heavy real usage and active troubleshooting rather than a static hype drop
  • Listed on the official Claude plugin marketplace
  • Multiple independent third-party reviews document hands-on week-long usage with specific praise and specific complaints

feedbacks

What People Are Saying

  • "the agent asks detailed questions before coding, then creates a CONTEXT.md"Devtalk forum review

  • "if you already know what you want, this cuts agent verbosity to near zero"Devtalk forum review

  • "great for isolated functions, but on complex UI components the test-first approach slows things down"Devtalk forum review

  • "extract every piece of context out of your head before you start prompting"X reply

  • "made me rewrite how I work with AI agents"Medium article

  • "allows iteration one to start at 90% accuracy, so you refine instead of clawing up from the bottom"Dev.to article

GitHub30.0K

volcengine/OpenViking

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

youtube
OpenViking: The Context Database AI Agents Actually Need

Market Signal

Why It Has Market Pull

OpenViking is a real, actively maintained open-source project built by Volcengine, ByteDance's cloud computing division, giving it credible backing rarely seen among AI-agent tooling. It has drawn genuine developer attention, reaching the number-one spot on GitHub's trending page and accumulating tens of thousands of stars, and independent demo videos and technical write-ups have followed its release.

  • Over 30,000 GitHub stars and 2,300+ forks
  • Built and maintained by Volcengine, ByteDance's cloud computing division
  • Reached #1 on GitHub's trending page in March 2026
  • Covered by independent outlets including MarkTechPost and FAUN.dev, plus multiple independent YouTube walkthroughs
  • Active issue tracker and versioned releases show ongoing maintenance

feedbacks

What People Are Saying

  • "openviking-server doctor doesn't report a config error"GitHub issue

  • "Fresh pip install produces dependency conflict warnings — is this expected and the setup OK?"GitHub issue

  • "reaches 52.08% task completion at 4.2M input tokens, versus 35.65% without the plugin"MarkTechPost coverage

  • "unifies the management of context (memory, resources, and skills) that agents need through a file system paradigm"GitHub README

  • "Independent third-party coverage is still thin."evidence gap

HF Spaces193 likes

MiniMax Music 3 Studio

MiniMax Music 3 Studio — diffusers demo Streams full songs from lyrics + a structured caption using the MiniMaxMusic3Pipeline diffusers port. The input surface is a single Suno-inspired custom gr.HTML composer (Simple ↔ Studio modes, section-tag chips, structured-caption fields per the official prompting guide) that drives Gradio events via trigger()/props.value; styling uses only theme CSS vars so it follows the Citrus theme natively. Weights: MiniMaxAI/MiniMax-Music3 AoTI kernels: diffusers-internal-dev/MiniMax-Music3-aoti (compiled on RTX Pro 6000, matching ZeroGPU hardware) Generation streams chunk by chunk with a configurable playback headroom. The 8B language-model stage runs eager on....

youtube
MiniMax-Music3: Open Music Generation Model Tested Locally

Market Signal

Why It Has Market Pull

MiniMax Music 3 is a genuine, actively developed release from a major, well-funded AI lab, publicly framed by Hugging Face staff as a state-of-the-art open-weight alternative to closed tools. Early hands-on reviews confirm it produces clean, structurally coherent full songs in English, though quality drops for non-English genres — real, specific feedback rather than hype.

  • 196 likes on Hugging Face with active updates as recent as August 13, 2026
  • Built by MiniMaxAI, a major, well-capitalized foundation-model lab
  • Publicly called state-of-the-art open-weight song generation by a Hugging Face staff member on release
  • Generates full songs up to five minutes from lyrics and a style description, a differentiated capability versus most short-clip music demos
  • Independent reviews confirm clean, coherent English-language output but flag weaker results for non-English genres

feedbacks

What People Are Saying

  • "MiniMax Music 3 is just out on Hugging Face... state of the art"X post

  • "Great work!"HF Space discussion

  • "seems to take everything I make and turn it into a rap song"user comment via independent coverage

  • "The audio was clean, structurally sound, and didn't fall apart over the whole track, which is a real technical achievement"independent review

  • "didn't match the genre's real characteristics at all"independent review

  • "Provide automatic copy of completed song/video files to permanent user storage"HF Space discussion

HF Spaces104 likes

OpenVuln

Find bugs in your repository with GLM This Space is built automatically from the root Dockerfile and serves the Vite application with nginx on port 7860. Optional Space build variables: VITEAPIBASEURL — API origin; defaults to https://openvuln.vulnhunter.pro. VITEGITHUBREPOURL — source repository linked from the interface. OpenVuln is a Hugging Face Space tagged with docker, region:us. It has 104 likes on Hugging Face.

youtube
China's New AI Broke the Bug Bounty (2,436 Findings)

Market Signal

Why It Has Market Pull

OpenVuln is a genuine security initiative from Z.ai (Zhipu AI), the lab behind the GLM model family, using GLM to automatically audit open-source repositories for vulnerabilities and coordinate responsible disclosure. It has real substance behind it — dozens of CVEs already assigned and thousands of findings logged across hundreds of projects — making it one of the more credible AI-for-security tools currently public.

  • Hosted under the verified Hugging Face organization for Z.ai (Zhipu AI), makers of the GLM model family
  • 105+ likes on Hugging Face and coverage across multiple security and tech outlets at launch
  • 53 vulnerabilities already disclosed with CVEs assigned, and 2,436 findings logged across 269 open-source projects, 1,097 rated critical or high severity
  • Released alongside GLM-5.3, described as Z.ai's strongest model yet for cybersecurity tasks, achieving 84.5% on the CyberGym vulnerability-finding benchmark
  • Security researchers have raised open questions about downstream misuse once model weights are public — a real, actively-discussed caveat rather than pure hype

feedbacks

What People Are Saying

  • "GLM-5.3 delivers substantial improvements in vulnerability discovery, exploit analysis, and complex multistep security tasks"Tech press coverage

  • "OpenVuln Launches VulnHunter AI for Public Open Source Vulnerability Intelligence"Tech press headline

  • "missing model cards and licenses [are] material omissions for researchers determining permitted uses and safeguards"Security researcher commentary

  • "free continuous security auditing for open-source projects"Tech press coverage

  • "Preparing GLM-5.3 for Open Release: A Responsible Path to Cyber Defense"Z.ai announcement (X)

  • "1,097 classified as critical or high severity"Tech press coverage

Sources

GitHub

immich-app/immich

111.7K

High performance self-hosted photo and video management solution.

harry0703/MoneyPrinterTurbo

110.2K

利用 AI 大模型和自动化工作流，根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

santifer/career-ops

65.6K

Open-source AI job search: scan job portals, evaluate listings with a structured A-F rubric into a 1.0-5.0 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)

mukul975/Anthropic-Cybersecurity-Skills

29.7K

817 structured cybersecurity skills for AI agents · Mapped to 6 frameworks: MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND, NIST AI RMF & MITRE F3 (Fight Fraud) · agentskills.io standard · Works with Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI & 20+ platforms · 29 security domains · Apache 2.0

jundot/omlx

19.8K

LLM inference server with continuous batching & SSD caching for Apple Silicon — managed from the macOS menu bar

Product Hunt

Clara AI SDR

437

Clara is an AI SDR that converts inbound website traffic into pipeline - engaging, qualifying, demoing, handling objections, and booking meetings in real time. No forms. No waiting. No extra headcount. 24/7 selling with seamless CRM integration.

Taku AI

399

AI is getting insanely powerful — and somehow harder to use. Taku fixes the last mile. It turns the best skills, agents, and workflows into real desktop apps anyone can run, remix, and make their own. Borrow setups from the pros, skip GitHub and setup, or just tell Taku what you want and watch it assemble the stack. The more you work, the more Taku learns what you need — and brings the right tools to you.

Superflow AI

258

Superflow turns the QA checklist you already use into a team of AI agents. They sweep every page, desktop and mobile, and pin every finding on the live site. Agents handle the black and white issues. Taste stays with you. Agents learn: rejected findings stop coming back, and misses become checks on future sites. Teams say they catch ~90% of what they found by hand. Works with Webflow, Framer, WordPress, Shopify, Next.js, Netlify + more.

Hubble

196

Medical records still live behind fax lines, phone trees, and portal logins nobody remembers. A patient verifies their identity once, and Hubble assembles their records from across their providers and returns them through one API, for your AI agent to use.

ElevenLabs MCP in Claude

160

Connect Claude to your ElevenLabs workspace to create and manage voice agents. Find existing agents, review their configuration, update prompts and voices, duplicate or delete agents.

AirBuddy 3

135

AirBuddy 3 brings over 150 new features and improvements, including a new design, powerful new ways to keep an eye on your devices, deeper media integration, revamped Magic Handoff, and many new options for automation and customization.

YC Launch

idler — push the frontier of your model

8

Hill-climb any benchmark, for any model, in any domain using our RL environments. idler · Summer 2025 · B2B Tags: Reinforcement Learning. Website: https://idler.ai/

Grip - Robotics for Waste Management

6

Robots that sort any waste and recover recyclables, learning from every grasp. Grip · Summer 2026 · Industrials Tags: Robotic Process Automation, Sustainability. Website: https://www.griprobotics.ai

Earendil | Autonomous Air Defence

Autonomous, combat-proven interceptors for defending against drone swarms. Earendil Robotics · Summer 2026 · Industrials Tags: Drones, Swarm Robotics, AR. Website: https://earendil.io

Hacker News

Mocktail – Free, open-source mock API server with a built-in dashboard

24 pts

Hi HN, I built the first version of Mocktail years ago. I recently came back to the project, and after a pretty substantial rebuild, v4 is now out. Mocktail is a free and open-source, self-hosted mock API server with a built-in dashboard and database, packaged as a single ~25 MB binary. You can run it locally or on your own infrastructure — no account or hosted service required. You can define endpoints and responses, generate realistic data per request, customize headers, status codes and laten... (24 points, 1 comments).

HarnessRouter: Unified interface for agent harnesses

9 pts

Hey HN! We are building HarnessRouter, a canonical API for running Codex, Claude Code, Hermes, and other managed agent harnesses as your product backend. Before building HarnessRouter, I used to build our own agent harness for our products. I tried LangGraph, agent SDKs from different vendors, pydantic, LLM tool use / function call, and so on. It's a very heavy lifting engineering effort, and I am disappointed about the agent deliveries compared to what Codex, CC can deliver. That changed my min... (9 points, 13 comments).

MCP app for Android, drive apps via AI (no root, PII redacted locally)

5 pts

Hi HN, author here. Android Remote Control MCP is an MCP server that runs directly on your Android phone (no root, no ADB, no computer in the middle) and lets an AI agent drive real apps the way a human would: it reads the screen through the accessibility tree, taps, types, scrolls, optionally also screenshots. I spent a lot of time optimizing tool usage and token consumption, and making it work not only via local harnesses but also via Claude.ai / Claude Desktop and chatgpt.com (if you have the... (5 points, 0 comments).

1667, a terminal UI for writing fiction with language models

37 pts

Hi HN. I built 1667 for my own fiction work and now use it each day. This probably has a limited audience. Maybe an audience of one... Why a terminal interface for story writing? I'm a dev. I like to use terminals for a lot of stuff. Most WebUIs feel off to me. That's the only reason. One thing that bothers me about writing in existing tools is that they don't fit the way I write. The mental model of my story is a tree. I try many takes usually continue with just one, but sometimes I want to try... (37 points, 93 comments).

Doberman: The AI watchdog that stops Claude from deleting your database

9 pts

Landing Page: https://www.trydoberman.dev/ The Problem: AI guardrails only protect against 89% of dangerous commands. This may seem pretty good until you think of it as a lock that lets 1 in 9 intruders in. What is Doberman: Doberman is a two layer security system wrapped around your LLM monitoring every input, output and tool call. The first layer is deterministic based on state of the art security guidelines, the second layer is dynamic, learning from your use and personal preferences. How is... (9 points, 5 comments).

HF Spaces

Qwen-Image-Edit-2511-LoRAs-Fast

2.6K likes

Demo of the Collection of Qwen Image Edit LoRAs Qwen-Image-Edit-2511-LoRAs-Fast is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 2624 likes on Hugging Face.

UGI Leaderboard

2.0K likes

Uncensored General Intelligence Leaderboard UGI Leaderboard is a Hugging Face Space tagged with docker, leaderboard, submission:manual, test:private, modality:text. It has 1991 likes on Hugging Face.

FLUX.2 Klein multi-LoRA

406 likes

Use multiple FLUX.2-Klein LoRAs Note: This space is experimental and may log image-uploads during certain periods for performance monitoring. Always comply with HF and model Terms of Service. Any stored images are automatically deleted after 7 days. FLUX.2 Klein multi-LoRA is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 406 likes on Hugging Face.

MiniMax H3 Turbo LoRA

226 likes

Video generation with a synchronized soundtrack MiniMax-H3 — unquantized, split across two Spaces Joint video and soundtrack out of a single denoising pass, at bfloat16 with no quantization anywhere. This Space is the denoising half: the 61.73 GiB transformer and the two autoencoders. The 62.14 GiB Qwen3-VL conditioner runs in qwen3vl-conditioner, which this Space calls over the gradio API for every request. The weights are the public MiniMaxAI/MiniMax-H3 diffusers checkpoint. MiniMax-H3 is 195.9 GiB in bfloat16 and a ZeroGPU Space is evicted at 150 GB of storage. An unquantized single Space is therefore impossible, which is why quantized demos of it run NVFP4 or float8 weights. Cut the Mini...

MiniMax-H3 Ultra Fast

125 likes

Ultra-fast local NVFP4 video + synchronized audio generation MiniMax-H3 Ultra Fast — local conditioner + pruned NVFP4 on Blackwell Joint video and synchronized sound from MiniMax-H3, rebuilt for a single 96 GB Blackwell ZeroGPU worker. | layer | optimization | |---|---| | Weights | 12.5 GB pruned NVFP4 transformer: 20.1B effective parameters instead of 33.1B/61.7 GiB BF16. | | Compute | Native CUDA 13 NVFP4 tensor-core GEMMs through comfy-kitchen; higher-precision norms, embeddings and output heads. | | Residency | Transformer, conditioner and both VAEs remain GPU-resident during generation—no layerwise CPU offload. | | Conditioner | Local 15.7 GB Qwen3-VL NVFP4-AWQ checkpoint containing onl...

Qwen3.8-27B Free Endpoint

64 likes

Guide to the free, rate-limited, OpenAI-compatible community endpoint for Qwen/Qwen3.8-27B (vision, tool calls, 262K context, dialable thinking), styled as a classic Mac desktop. Qwen3.8-27B Free Endpoint is a Hugging Face Space tagged with static, region:us. It has 64 likes on Hugging Face.