This is the most slept-on repo in AI coding right now
Addy Osmani (Google) packed senior engineering into 25 skills for your agent
117,000 stars, #1 repo of the day, and most devs still let Claude freestyle
it installs into 70+ agents with one command:
npx skills add addyosmani/agent-skills
the 6 rules inside:
• rule 1 -> /spec
"spec before code"
vague idea -> PRD -> then the first line of code
• rule 2 -> /plan
"small, atomic tasks"
spec -> tasks with acceptance criteria -> dependency order
• rule 3 -> /build
"one slice at a time"
implement -> test -> verify -> commit -> next slice
• rule 4 -> /test
"tests are proof"
red -> green -> refactor. "seems right" is never enough
• rule 5 -> /review
"improve code health"
five-axis review, ~100 lines per change, labeled nits
• rule 6 -> /ship
"faster is safer"
feature flag -> staged rollout -> rollback ready
the best part: every skill has an anti-rationalization table
agent says "I'll add tests later" -> the skill already has the rebuttal
your agent doesn't need a smarter model. it needs a senior engineer's habits
repo is free, link below ↓
https://t.co/goRQUDp7Rp
This is the most slept-on repo in AI coding right now
Addy Osmani (Google) packed senior engineering into 25 skills for your agent
117,000 stars, #1 repo of the day, and most devs still let Claude freestyle
it installs into 70+ agents with one command:
npx skills add addyosmani/agent-skills
the 6 rules inside:
• rule 1 -> /spec
"spec before code"
vague idea -> PRD -> then the first line of code
• rule 2 -> /plan
"small, atomic tasks"
spec -> tasks with acceptance criteria -> dependency order
• rule 3 -> /build
"one slice at a time"
implement -> test -> verify -> commit -> next slice
• rule 4 -> /test
"tests are proof"
red -> green -> refactor. "seems right" is never enough
• rule 5 -> /review
"improve code health"
five-axis review, ~100 lines per change, labeled nits
• rule 6 -> /ship
"faster is safer"
feature flag -> staged rollout -> rollback ready
the best part: every skill has an anti-rationalization table
agent says "I'll add tests later" -> the skill already has the rebuttal
your agent doesn't need a smarter model. it needs a senior engineer's habits
repo is free, link below ↓
https://t.co/goRQUDp7Rp
Stanford AI engineering course: "Anyone can build an AI agent in 60 minutes"
Prompt → Agent → Automation → Revenue Stanford
just released a course on building AI agents from scratch
00:00 - Build your first AI agent
48:17 - Create agents without coding
54:39 - Make $100K+ per month with agents
While you scroll, someone else is learning Anthropic's $750,000 skill
This free course is better than most paid AI agent courses
Bookmark and watch it today
Then read the article below
This is f*cking insane. Anthropic CEO Dario Amodei was asked what a kid finishing high school should do now. His answer:
"AI can do entry-level law work. It can do entry-level finance. It can do entry-level consulting
We can't deny that it will also eliminate jobs, and probably a large number of them"
In a few minutes he names which jobs go first and the one skill that keeps you in the game
His own prediction: 50 percent of entry-level white-collar jobs disrupted in the next 1 to 5 years
Prompts → Agents → AI Coworkers → Entry-Level Work Done By AI
Farming to factories took generations. This time AI hits every step of a career at once
His advice fits in one line: "Of course, learn to use AI"
The people who start this year won't be competing with AI, they'll be the ones running it
Everything below is free. Start today:
Anthropic - Prompt Engineering Interactive Tutorial: the fastest way to learn to direct a model
https://t.co/vsqcKsskwn
Anthropic Courses: from your first prompt to running agents
https://t.co/rzfgfmSIKF
Anthropic Cookbook: real task recipes you can copy today
https://t.co/0ijufv3kKY
Microsoft - Generative AI for Beginners: 21 lessons, no degree required
https://t.co/l02hgPl6Gn
Microsoft - AI Agents for Beginners: build agents that do the entry-level work for you
https://t.co/MPhRrtaRBy
Hugging Face - Agents Course: free and certified, start to finish
https://t.co/loj9B7jgkX
Google Cloud - Generative AI: Google's own hands-on notebooks
https://t.co/A61I6WYtMM
OpenAI Cookbook: code for every common task
https://t.co/qfhNnYTGuo
https://t.co/6OUUA2MJ7b - Prompt Engineering Guide: the reference everyone links to
https://t.co/jLYld9cDR3
Andrej Karpathy - Neural Networks Zero to Hero: understand what is under the hood
https://t.co/YwL0IURo1V
CrewAI: run a small team of AI workers
https://t.co/cwrDl2xFvW
Microsoft - AutoGen: put several agents on one job
https://t.co/W7QINWf9qJ
Awesome LLM Apps: hundreds of real builds to learn from
https://t.co/PX2XHrLKWv
https://t.co/LPe5lLx25N - Short Courses: free 1-hour courses from the people who teach this
https://t.co/jKa6NuhCqH
Total cost: $0
You don't need another year of tuition to learn this. You need to start telling AI what to do
95% will graduate and compete with AI for the same entry-level job
The 5% who learn to run it this year will manage the agents doing that job
Save this and send it to someone graduating this year
Then read the full Graph Engineering guide below ↓
Alibaba, NVIDIA and Google just open-sourced the tools their own engineers use to keep AI agents in check
67,120 stars between them. all free
this is the stuff big tech used to keep behind the firewall:
1. open-code-review by Alibaba -> the reviewer (50.3K stars, #1 repo of the day)
Alibaba's internal AI code reviewer for 2 years. tens of thousands of devs, millions of defects caught. on their benchmark it beat Claude Code on precision using ~1/9 of the tokens
↳ https://t.co/zqgu88eoc5
2. auto-ontology by NVIDIA -> the fact-checker (12.5K, +5.8K this week)
your agent writes SQL, gets a revenue number, sounds confident. it's wrong. this gives agents a map of what your data actually means, checks every query against certified definitions and returns provenance with each answer. plugs in over MCP
↳ https://t.co/n8RGThsuXo
3. modern-web-guidance by Google -> the senior frontend dev (4.4K, +1,875 today)
your agent writes 2019 JavaScript because that's what it trained on. this skill feeds it modern web platform patterns instead. stewarded by the Chrome, Edge, Firefox and MDN teams
↳ https://t.co/FXDsoe39hI
here's how to set them up:
1. npm install -g @alibaba-group/open-code-review && ocr config provider
2. npx modern-web-guidance@latest install
3. git clone https://t.co/dXjcRIKNF9 && cd auto-ontology && cp .env.example .env && docker compose up -d --build
then:
"ocr review" before every PR → "npx modern-web-guidance@latest legacy-scan ." on your frontend → point your agent at auto-ontology's MCP server before it touches your data
the best agent tooling isn't coming from startups anymore. it's leaking out of big tech
which one are you installing first? ↓
REA is most dangerous thing on the internet right now...
It lets your AI agent reverse engineer any app, down to the binary
If you use it correctly, you can rebuild any feature you pay for, for FREE
In this article, I show you how: https://t.co/3P1FNXiXzl
Cloudflare, OpenAI and shadcn all shipped the same answer: let a second agent audit the first one.
91,749 stars combined
first understand what each one does:
1. security-audit-skill by Cloudflare -> the pentester (69.9K stars, +11K this week)
the skill that seeded Cloudflare's own vulnerability harness. isolated agents hunt bugs, then a fresh agent tries to disprove every finding before it hits the report
↳ https://t.co/FzjpbIm614
2. codex-security by OpenAI -> the scanner (11.4K)
finds, validates and patches vulnerabilities. scan the whole repo or just your diff, run it in CI, push findings to Linear, Jira or GitHub
↳ https://t.co/a4YoTEMAPr
3. improve by shadcn -> the senior reviewer (10.4K)
your smartest model audits the codebase and writes plans. it never touches the code, cheaper models execute
↳ https://t.co/Z71ReUoyPQ
here's how to set them up:
1. npx skills add https://t.co/fqWR83Tlkm --skill security-audit
2. npm install -g @openai/codex-security && cs login
3. npx skills add shadcn/improve
then before your next PR:
"security audit this codebase" → cs scan . --diff origin/main → /improve branch
your agent ships fast. make the second one ship it safe
what does your review step look like? drop it below ↓
REA is most dangerous thing on the internet right now...
It lets your AI agent reverse engineer any app, down to the binary
If you use it correctly, you can rebuild any feature you pay for, for FREE
In this article, I show you how: https://t.co/3P1FNXiXzl
Everyone is upgrading the model
the people shipping are upgrading the harness
agent = model + harness. the model is rented, the harness is yours
3 repos trending on GitHub this week, 390,692 stars combined. together they're the harness most people never build
first understand what each one does:
1. agent-browser -> the eyes (49K stars)
Vercel's browser CLI for agents. it reads pages as an accessibility tree, not screenshots, so every step is one short shell call
↳ https://t.co/tBiyvkAeQE
2. agent-desktop -> the hands (21.2K)
same idea for your whole computer. operates real desktop apps on macOS, Windows and Linux, and there's no model inside, the loop stays in your agent
↳ https://t.co/LGpCya1Fre
3. skills by Matt Pocock -> the habits (321K)
the skills he uses every day for real engineering, not vibe coding. small, composable, works with any model
↳ https://t.co/rLsacpWrvr
here's how to put them on:
1. npm install -g agent-browser && agent-browser install
2. npx skills add vercel-labs/agent-browser
3. npm install -g agent-desktop
4. claude plugin install mattpocock-skills@claude-plugins-official
then paste this into Claude Code:
"use agent-browser to open my app, click through signup and tell me everything that breaks"
same model. different agent
what's in your harness? drop repo links below ↓
REA is most dangerous thing on the internet right now...
It lets your AI agent reverse engineer any app, down to the binary
If you use it correctly, you can rebuild any feature you pay for, for FREE
In this article, I show you how: https://t.co/3P1FNXiXzl
Claude Code forgets everything the second you close the terminal
7 free repos fix that. 230,706 stars. Obsidian becomes its brain, and Claude writes the notes itself
in the right order:
1. claude-obsidian -> the brain
drop any source in, Claude reads it, files it and links it. self-organizing (16.3K stars)
↳ https://t.co/TddZOisg2k
2. obsidian-skills -> the hands
skills from Obsidian's CEO himself. teaches your agent the CLI and every vault format (51.3K)
↳ https://t.co/jtwKWuZLdq
3. karpathy-llm-wiki -> the method
Karpathy's LLM wiki idea packed as one skill. raw sources in, living wiki out (2.7K)
↳ https://t.co/CyaVrRf9Z3
4. claude-mem -> the memory
captures every session and feeds it back next time. never re-explain your project again (115K)
↳ https://t.co/avwnfw1un1
5. mcp-obsidian -> the bridge
an MCP server that lets any agent read and edit your vault (4.5K)
↳ https://t.co/lGNuZoIb6b
6. graphiti -> the graph
turns notes into a real-time knowledge graph your agent can query (32.7K)
↳ https://t.co/VvhWN9z2XA
7. obsidian-copilot -> the cockpit
run Claude Code and Codex right inside Obsidian (8K)
↳ https://t.co/jMw8FQZbqJ
start with 1 + 2. one vault, one prompt: "read raw/, update the wiki, link everything"
a month later you have a second brain you never typed a single note into
and if you want to know how people turn that vault into $2,100 a month ↓
REA is most dangerous thing on the internet right now...
It lets your AI agent reverse engineer any app, down to the binary
If you use it correctly, you can rebuild any feature you pay for, for FREE
In this article, I show you how: https://t.co/3P1FNXiXzl
closed-source just became optional...
8 free repos turn Claude Code into a full reverse-engineering lab. 662,810 stars between them
point it at an app, APK or binary → it maps how the thing works → then builds you your own version
in the right order:
1. rea -> the map
one MCP for apps, bundles and native binaries. every finding saved as evidence (443K stars)
↳ https://t.co/kvkjyJQahu
2. GhidraMCP -> the decompiler
Claude drives the NSA's Ghidra over MCP (11.8K)
↳ https://t.co/RxLEDngUKM
3. ImHex -> the raw bytes
the hex editor reverse engineers actually use (55.8K)
↳ https://t.co/jXZNJyHrOT
4. jadx -> Android
Dex → readable Java (51.7K)
↳ https://t.co/Z7aDBu2Wus
5. android-reverse-engineering-skill -> the APK playbook
a Claude Code skill built just for APKs (8.7K)
↳ https://t.co/B4Ni57PfaB
6. frida -> runtime
hook a live app and watch what it actually does (22.7K)
↳ https://t.co/atlDwlQUej
7. mitmproxy -> the network
intercept HTTPS, map the hidden API (46.7K)
↳ https://t.co/mkq4frSyYl
8. craft-anything -> the payoff
studies how it works, then builds your own version (22.3K)
↳ https://t.co/Z57P7okE0k
1-7 tell you how it works. 8 builds yours
the part most people skip: clean room. run it on software you own or are licensed to inspect, and never ship copied code
whoever learns this stack stops paying for tools and starts shipping the alternatives
here's what it looks like when someone ships 10 apps with it ↓
REA is the most powerful thing on the internet right now
it lets your AI agent reverse-engineer any app from scratch
i connected REA to Opus 5.5 and reverse-engineered paid 10 apps that gained 13K+ installs
in this article, I reveal the full workflow: https://t.co/NB6UZqfFh9