Tools Bench.

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

Last Brew Time: Sep 3, 2026, 11:04 AM PT

Insight

The loudest builders in this run aren't shipping new software so much as packaging their own workflow opinions as install-in-a-command agent skills, with infrastructure companies increasingly treating the AI agent itself as the paying customer

01.SKILL PACKS THAT BUNDLE A DEVELOPER'S OWN PLAYBOOK

02.PERSONA RULE-FILES THAT RESHAPE HOW AN AGENT CODES

03.INFRASTRUCTURE BUILT TO SERVE AN AGENT AS ITS OWN CUSTOMER

Featured

GitHub281.2K

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 turns Claude Code into a full software-development partner by bundling proven workflows—structured brainstorming, test-driven development, and systematic debugging—into one install. In under a year it has become one of the most widely adopted open-source AI coding frameworks, praised by respected developers for changing how they work day to day, though some engineers find its multi-step process heavier than they need for small tasks.

  • 281,260 GitHub stars and 25,191 forks as of September 2026, up from zero at its October 2025 creation
  • 349 open GitHub issues, showing an active, engaged user base rather than a dormant repo
  • Reported adoption by roughly 50,000 developers within the first few months of release
  • Datasette creator Simon Willison publicly called its creator one of the most creative users of coding agents he knows
  • Repeated fresh coverage into 2026 (Medium, mcp.directory, Termdock) months after launch shows sustained rather than one-off interest

feedbacks

What People Are Saying

  • "Superpowers, the most installed skills framework for Claude Code. 50,000 developers adopted Superpowers in the first few months."X post (Russell Fradin)

  • "one of the most creative users of coding agents [I know]"Simon Willison commentary

  • "shifts vibe coding into agentic engineering"HN discussion (paraphrased consensus)

  • "over-engineering for small tasks... bloated"HN discussion (top criticism)

  • "I Gave Claude Code a Brain. It's Called Superpowers — and It Has 150,000 GitHub Stars for a Reason"Medium review headline

  • "Superpowers for Claude Code: Still Worth It in 2026?"mcp.directory blog headline

GitHub240.7K

NousResearch/hermes-agent

The agent that grows with you

youtube
Hermes Agent First Look - Nous Research's Self-Improving AI Agent (Quickstart Walkthrough)

Market Signal

Why It Has Market Pull

Hermes Agent is a self-improving personal AI assistant built by Nous Research that runs from a terminal, desktop app, or chat platforms like Telegram, Discord, and Slack, learning new skills from each session and remembering them across conversations. The project has scaled from an open-source release into a venture-backed business reportedly raising over $75 million at a $1.5 billion valuation, with paid hosted plans already on the market. It has become one of the most widely adopted open agent projects this year, though its rapid rise has also drawn scrutiny over attribution and safe-usage practices.

  • 240,700+ GitHub stars and 49,300+ forks as of September 2026, up from about 214,000 stars in July 2026
  • Over 5 million pulls on its official Docker Hub image
  • Raised a $50 million Series A led by Paradigm and reportedly raising $75M+ more at a $1.5B valuation, with backing from Robot Ventures, USV, North Island Ventures, Delphi Ventures, OSS Capital, and Balaji Srinivasan
  • Active same-day development commits, plus paid hosted plans from $20 to $200/month
  • Also drew negative press: a hacker used its 'YOLO mode' for a real post-exploitation attack, and a public GitHub dispute over similarity to EvoMap's Evolver

feedbacks

What People Are Saying

  • "I've been using the NousResearch Hermes agent for the past couple of weeks and have to say it's been really good"HN comment (rnxrx)

  • "Nous Research's Hermes appears derivative of (or heavily inspired by) EvoMap's Evolver, but without attribution."HN comment (0hijinks)

  • "Hermes set up locally with Telegram doesn't respond despite following all instructions"GitHub issue #853

  • "Telegram pairing list shows non-approvable code in CLI"GitHub issue #46580

  • "Agent cannot read Telegram group chat history via t.me/c/ URLs (HTTP 403)"GitHub issue #10020

  • "Probe still failing... Index is 49.5h old (limit 26h)"GitHub issue, skills-index-watchdog

  • "a serious scope-fidelity failure where the agent represented a control-plane skeleton as a completed execution model"user complaint roundup, aiagentstore.ai

GitHub173.5K

anthropics/skills

Public repository for Agent Skills

youtube
Claude Skills Tutorial (2026): Build, Run, and Share

Market Signal

Why It Has Market Pull

This is Anthropic's own public repository of Agent Skills, the packaged instructions and resources that let Claude and other AI agents complete specialized tasks reliably. It has grown into the reference implementation for what's become an open cross-industry standard, adopted well beyond Claude by tools like GitHub Copilot, Cursor, and OpenAI Codex, with major companies contributing their own official skill packs.

  • 173,573 GitHub stars and 20,594 forks, actively updated as of today
  • Star count has roughly doubled since February 2026, from about 149k
  • Adopted as an open standard (agentskills.io) by ~40 other agent clients including GitHub Copilot, Cursor, OpenAI Codex, and Gemini CLI
  • Partner skill directory now includes Atlassian, Canva, Cloudflare, Figma, Notion, Ramp, and Sentry
  • Enterprise admin controls added alongside the open-standard rollout

feedbacks

What People Are Saying

  • "Claude Skills are awesome, maybe a bigger deal than MCP."Hacker News / Simon Willison

  • "Skills teach Claude how to complete specific tasks in a repeatable way, whether that's creating documents with your company's brand guidelines or automating personal tasks."Repository documentation

  • "The repo has roughly doubled to ~149k GitHub stars since February 2026."Independent coverage (generativeprogrammer.com)

  • "Adopted well beyond Claude — GitHub Copilot, VS Code, Cursor, OpenAI Codex, Gemini CLI, Goose, OpenCode, and ~40 other clients."Industry coverage

  • "Anthropic added enterprise admin controls plus a partner skills directory."Industry coverage

Product Hunt453

Monid

We connect your AI agent to 1,800+ APIs without any subscriptions. SEO, lead gen, video/music generation, social media, stocks, market trends, on-chain data, competitor tracking, sentiment analysis, all unlocked with one key.

youtube
Monid AI

Market Signal

Why It Has Market Pull

Monid gives an AI agent one key to reach more than 1,800 paid data and content tools, covering search, SEO, lead generation, market and social data, and image, video, audio and 3D generation, so builders stop signing up for a new subscription every time an agent needs a new capability. It landed as the number-one launch of the day on Product Hunt and has already backed more than 4 million agent tool calls, with a $2.1 million pre-seed round behind it.

  • Ranked #1 product of the day on Product Hunt with 455 upvotes
  • Raised $2.1M pre-seed from 1984 Ventures, Llama Ventures, Untapped Capital, and Founders Inc
  • Agent transactions through the platform have passed 4 million
  • Catalog has grown from roughly 200 to more than 1,800 connected APIs and tools
  • Ships as an MCP server, an agent skill, a CLI, and an HTTP API for flexible integration

feedbacks

What People Are Saying

  • "One key for 1,800 APIs instead of separately signing up, testing, and billing for each one."Product Hunt review (Gal Dayan)

  • "A bad retry loop or stuck agent could rack up a real bill."Product Hunt review (Gal Dayan)

  • "Been using it daily — it's saved me at least 15 hours of searching and setting up tools, plus a few hundred dollars."Product Hunt comment (Aurora Zhang)

  • "This is such a relatable problem — hunting for API providers, signing up, testing with real data, then finding out the provider is inadequate."Product Hunt comment (Ryan Hoover)

  • "How do you handle rate limits across all these different providers?"Product Hunt comment thread

  • "Differs from pre-wired integration platforms like Merge by letting agents discover and pay for tools at runtime."Product coverage (MOGE/AIDiveForge)

Product Hunt141

Claude Fable 5.1

Claude’s most advanced models for coding and knowledge work. Their research capabilities also offer an early glimpse of how AI models will contribute to scientific progress.

youtube
Introducing Claude Fable 5.1

Market Signal

Why It Has Market Pull

Claude Fable 5.1 is Anthropic's newest flagship model for coding, agentic work, and knowledge tasks, priced roughly 25-45% cheaper than its predecessor for typical and heavily agentic workloads while cutting cybersecurity false positives by about 60%. The release drew coverage across MacRumors, 9to5Mac, and VentureBeat and sparked active discussion among developers already running it in production agent workflows.

  • 141 upvotes on the Product Hunt launch listing
  • Roughly 25% cheaper for typical workloads, up to 45% cheaper for highly agentic work vs Fable 5
  • About 60% fewer cybersecurity false positives in Claude Code, per Anthropic
  • Covered by MacRumors, 9to5Mac, VentureBeat, and Hacker News on its September 1, 2026 release

feedbacks

What People Are Saying

  • "i'm seeing things about fable i couldn't believe."Product Hunt comment (fmerian)

  • "Cache reads got cheaper, so typical workloads cost about 25% less than Fable 5"Product Hunt comment (Hunter KP)

  • "Better models are great until you actually run agents all day and see the bill."Product Hunt comment (Ben Cohen)

  • "wanted to know how it handles messy real world projects where the requirements keep changing"Product Hunt comment (Rajesh Kumar)

  • "reported hitting session quota limits quickly when using the highest effort mode intensively"Product Hunt comment (André J)

  • "demoed generating a house design, render, and cinematic walkthrough from a single property photo"Product Hunt comment (KP)

Sources

GitHub

affaan-m/ECC

246.9K

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

mattpocock/skills

246.9K

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

f/prompts.chat

168.9K

f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy.

DietrichGebert/ponytail

123.1K

Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.

JuliusBrussee/caveman

103.0K

🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman

Product Hunt

Browzer

424

Connect your GitHub repo and Browzer drafts your docs, guides, changelogs, cookbooks, quickstarts, and SEO & AEO friendly blog posts, and then heals them on every merge. We automate all technical output of a DevRel so that DevRel teams can focus more on community, growth & events.

Articos

365

Articos helps SaaS teams, agencies, and founders test messaging, positioning, and landing pages against simulated personas matched to your ICP. Real audience signal in under 30 minutes. Peer-reviewed at 86% human accuracy vs Baymard and Nielsen Norman.

CleanShot 5.0 with Studio Mode

260

CleanShot is the definitive Mac-native toolkit for screenshots, recordings and collaboration. The new Studio Mode is all you need to turn a simple screen recording into a polished, professional video — with the quality you expect from CleanShot. Available now on all plans. No subscription required.

Dial

221

Your agent can deploy in seconds but cannot get a phone number. Dial fixes that - one API call and it gets a real one in about 10 seconds. It places and receives voice calls, sends and receives SMS in 200+ countries, and messages over iMessage. It can even read inbound verification codes, so phone-gated signups stop being a dead end. Works via REST API, CLI, MCP servers, and SDKs. Free credit on signup, no card.

OpenClaw 2.0

207

OpenClaw 2.0 simplifies setup by detecting ChatGPT or Claude keys automatically and lets configuration happen through chats with the agent itself. It adds multiplayer team features for sharing sessions, refreshed browser tools for better tracking, and memory upgrades, all while running personal AI helpers locally on devices for tasks like browser control and file management.

Dyson CameraJet

141

The Dyson CameraJet combines sonic brushing with automated water flossing. An onboard macro camera detects interdental gaps at 28 fps to fire targeted conical bursts of mouthwash while you brush, removing the need for a separate flossing step.

YC Launch

Hacker News

Aura – a Rust agent that investigates and fixes production incidents

24 pts

We run a SaaS that handles petabytes of data. Our SRE team experimented with using claude, openclaw, langchain, etc. within our incident response workflows. We struggled with overflowing context, lethal trifecta vectors, hallucinations, and burned a lot of frontier tokens mostly on easy work. Approval fatigue was a challenge, and we drew a hard line at relaxing permissions in production. Long story short, we built and open-sourced AURA, a Rust-based harness specifically designed for the type of... (24 points, 4 comments).

I built my first MCP to manage Google Ads

13 pts

My co-founder and I tried running Google Ads for our Shopify store, but we could not get a positive ROAS (we spent 150USD on a single conversion). As we also could not afford an agency, we turned to AI (IMO Codex >>> Claude Code). We found that Google’s Ads MCP could only read data, but it could not make changes. So we built our own hosted MCP with read and write access. It can inspect performance, find wasted spend, create or update campaigns, and work with Google Analytics and Tag Manager. (13 points, 10 comments).

Ardent, a code-first agent for non-engineering work

7 pts

I’m Nate, the founder of Ardent. We just shipped our public beta, and we’d love your thoughts! Ardent is an agent running in a desktop (Electron) app built to help with knowledge work, designed for less-technical people outside engineering. Think Codex or Claude Cowork, but built around collaboration and customization. I know, I know, it’s yet another agent harness! Ardent is a little different – it leans heavily on codegen to solve problems. Most agents are basically just a bag of tools and a w... (7 points, 2 comments).

A Context Registry for AI coding agents

6 pts

Hi HN. We built an API context registry to help coding agents (like Claude Code) generate production-ready API integration code without blowing through token limits. We build a lot of API integrations. In our experience, most coding agents write basic client calls fine, but consistently stumble on details that make code shippable, like idempotent retries, rate-limiting and Auth token management. We tried all the existing approaches of injecting context into coding sessions: - Markdown dumps deli... (6 points, 1 comments).

Flawd is mutation testing for the AI era

4 pts

Hey, all! Flawd is a mutation testing tool that can target five languages (Python, JS, TS, Go, Rust) and it runs locally as a single binary on your machine. Importantly (for many), your code never leaves your machine and all processing takes place directly on your dev box or CI runner. Mutation testing is a process in which faults (known as mutants) are intentionally injected into your codebase, tests are run and any faults which are not detected by your tests are known as survivors (or survivin... (4 points, 5 comments).

A searchable, timestamped index of 1,124 AI Engineer talks

9 pts

I love the videos and talks that appear on my feed ever day from the AI Engineer conferences. Unfortunately, there are so many of them it's often hard to find the time to keep up, or to watch what I'd like. So I built aietalks.com, a searchable written index and archive of all the talks on the AI Engineer Youtube channel. We currently cover 1,124 talks. The individual talk pages offer tl;dr summaries, key quotes and other summarised information from the videos. Search and packs allow you to find... (9 points, 2 comments).

HF Spaces

Qwen-Image-Edit-2511-LoRAs-Fast

2.7K 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 2713 likes on Hugging Face.

Wan2.2 14B Fast Preview

1.7K likes

generate a video from an image with a text prompt Wan2.2 14B Fast Preview is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 1712 likes on Hugging Face.

FLUX.2 Klein multi-LoRA

505 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 505 likes on Hugging Face.

Microduck Sandbox

373 likes

Real trained RL policies for the Microduck robot, running fully in the browser: MuJoCo compiled to WebAssembly steps the physics, onnxruntime-web runs the policy network at 50 Hz. No server, no backend. Two locomotion variants of the same robot are included: legs (walking, the default) and rollers (the wheeled skating variant). Press M (or hold D-pad up ~1 s on a gamepad) to switch; the roller model, meshes and policies are lazy-loaded on the first switch. | Mode | Checkpoint | What it does | |--------|-----------|--------------| | Run (legs) | BESTalphawalking.onnx | Velocity-tracking locomotion (arrows / WASD to steer) | | Sit | BESTalphasitstand.onnx | Sits down on its hull, stands back u...

MiniMax H3 Turbo LoRA

364 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...

SenseNova-U1.5-8B-MoT

108 likes

Unified text-to-image and image editing model Text-to-image and image editing demo for sensenova/SenseNova-U1.5-8B-MoT, a natively unified multimodal model (18B params, bf16) built on the NEO-unify architecture. Leave the image upload empty for text-to-image generation, or upload one or more images and write an edit instruction for image editing. Advanced options exposes denoising steps, guidance scale, timestep shift, image guidance (editing) and the seed. The step count defaults to 28 rather than the model card's 50: a fixed-seed A/B found 28 keeps composition, prompt adherence and text rendering intact — losing only some micro-texture in landscape and skin, and nothing measurable when edi...

AI Tools — September 3, 2026 Edition | Agentic Brew