Agent Skills, Rules, Subagents: Explained!

Learn to use Anthropic's Claude Agent SDK (formerly Claude Code SDK) for AI-powered development workflows! https://platform.claude.com/docs/en/agent-sdk/overview https://x.com/trq212 **AI Summary** This workshop by Thariq Shihipar (Anthropic) details the architecture and implementation of the **Claude Agent SDK**. The session moves from high-level theory—defining "agents" as autonomous systems that manage their own context and trajectory—to a live-coding demonstration. Shihipar builds an agent "Harness" from scratch, implementing the core **Agent Loop** (Context Thought Action Observation), integrating the **Bash tool** for general computer use, and demonstrating **Context Engineering** via the file system to maintain state across long tasks. **Timestamps** 00:00 Introduction: Agenda and the "Agent" definition 05:15 The "Harness" concept: Tools, Prompts, and Skills 10:10 Live Coding Setup: Initializing the Agent class and environment 15:45 implementing the "Think" step: Getting the model to reason before acting 25:20 The Agent Loop: connecting `act`, `observe`, and `loop` 33:10 Tool Execution: Handling XML parsing and tool inputs 42:00 The "Bash" Tool: Giving the agent command line access 49:30 Safety & Permissions: "ReadOnly" vs "ReadWrite" file access 58:15 Context Engineering: Using `ls` and `cat` to build dynamic context 01:05:00 The "Monitor": Viewing the agent's thought process in real-time 01:12:45 Handling "Stuck" States: Feedback loops and error correction 01:21:20 Multi-turn Complex Tasks: Building a "Research Agent" demo 01:35:10 Refactoring patterns: "Hooks" and deterministic overrides 01:48:39 Q&A: Reproducibility, helper scripts, and non-determinism 01:50:31 Q&A: Strategies for massive codebases (50M+ lines) 01:52:00 Closing remarks and future SDK roadmap * **Evolution of AI Capabilities:** Shihipar argues we are shifting from **LLM Features** (categorization, single turn) to **Workflows** (structured, multi-step chains like RAG) to **Agents**. He defines agents as systems that *"build their own context, decide their own trajectories, and work very autonomously"* rather than following a rigid pipeline. * **The Claude Agent SDK Architecture:** The SDK is built directly on top of **Claude Code** because Anthropic found they were *"rebuilding the same parts over and over again"* for internal tools. * **The Harness:** A robust agent requires more than just a model; it needs a "Harness" containing Tools, Prompts, a **File System**, Skills, Sub-agents, and Memory. * **Opinionated Design:** The SDK bakes in lessons from deploying Claude Code, specifically the "opinion" that general computer use (Bash) is often superior to bespoke tools. * **The Power of the Bash Tool:** A key technical insight is that the **Bash tool** is often the most powerful tool for an agent. Instead of building custom tools for every action (e.g., a specific API wrapper for a file conversion), giving the agent access to the shell allows it to use existing software (like `ffmpeg`, `grep`, or `git`) to solve problems flexibly, similar to how a human developer works. * **Context Engineering:** Shihipar introduces the concept of **Context Engineering** via the file system. Instead of just "Prompt Engineering," the agent uses the file system to manage its state and context. * **Files as Memory:** The agent can write to files to "remember" things or create its own documentation (e.g., `CLAUDE.md`) to ground future actions. * **Verification:** The file system serves as a ground truth for the agent to verify its work (e.g., checking if a file was actually created). * **The Agent Loop & Intuition:** Building a successful agent loop is described as *"kind of an art or intuition"*. The loop generally follows a **Gather Context Take Action Verify Work** cycle. Shihipar emphasizes that this loop allows the agent to self-correct, a capability missing from rigid workflows. * **Strategies for Determinism (Hooks):** During the Q&A, a technique for controlling agent behavior is discussed: **Hooks**. * If an agent hallucinates or skips a step (e.g., guessing a Pokemon stat instead of checking a script), a hook can intercept the response and inject feedback: *"Please make sure you write a script, please make sure you read this data."* * This enforces rules like "read before you write" without retraining the model. * **Scaling to Large Codebases:** For massive codebases (50M+ lines), standard tools like `grep` or basic context window stuffing fail. * **Semantic Search Limitations:** Shihipar notes that while semantic search is a common solution, it is *"brittle"* because the model isn't trained on the specific semantic index. * **Solution:** He recommends good **"Claude MD"** files (context files) and starting the agent in a specific subdirectory to limit scope, rather than trying to index the entire 50M lines at once.

Visit sintra.ai/leonvanzyl or use my promo code leonvanzyl at checkout to receive limited time offer of 72% off on all plans 🚀 Free resources: source code, prompts, workflows + peer discussion https://skool.com/leonvanzyl 🧪 Agentic Labs: AI coding courses, live Q&A, weekly builder challenges, direct access to me + a serious builder community https://skool.com/agentic-labs OpenCode is the open source coding agent that is quickly becoming a serious alternative to Claude Code. In this opencode tutorial, you will learn how to install OpenCode, connect free AI models through OpenCode Zen or premium providers like OpenAI, and master advanced features including agent skills, sub agents, MCP servers, and background agents. ⏱️ CHAPTERS: 00:00 - OpenCode intro and overview 00:39 - Installing OpenCode in seconds 01:25 - Connecting providers and OpenCode Zen 02:20 - Free AI coding models walkthrough 03:46 - Scaffolding a Next js project 04:55 - Subscriptions vs API keys 06:00 - Connecting OpenAI and GPT 5 07:30 - Installing agent skills from skills sh 10:30 - AGENTS md memory file setup 13:00 - Sub agents and context window protection 15:30 - Design system for consistent UI 15:50 - Sponsor message 18:00 - Building the AI Project Planner app 22:30 - Saving plans for future sessions 24:30 - Creating custom sub agents 28:30 - Playwright MCP server setup 30:30 - Auto testing the app with browser MCP Business & sponsorship enquiries: leon.vanzyl@gmail.com #opencode #agenticcoding

In this episode, I sit down with Alex Finn for a full, screen-shared walkthrough of Hermes Desktop, the new desktop home for the Hermes AI agent. I open with a clear challenge: by the end, sell me on installing Hermes Desktop, show me real ways to make money and stay productive, and explain his move from OpenClaw. Alex tours every major surface — sessions, profiles, artifacts, skills, cron jobs, and sub-agents — and shares money-saving tactics at each step. We close on the idea that matters most to me: aiming these agents at other people's challenges as the clearest path to real value. Timestamps 00:00 – Intro 04:04 – Sessions and Context Management 06:10 – Profiles Explained 08:49 – Model-Based vs Role-Based Profiles 12:58 – Artifacts as a Second Brain 14:32 – Why Alex Switched From OpenClaw 17:32 – Skills, Tools, and Tool Sets 19:19 – Messaging and Cron Setup 21:44 – Reverse Prompting and the Brain Dump 28:09 – Sub-Agents vs Profiles 32:12 – Putting It Together: Solving Challenges 32:38 – The Daily Business Opportunity Scan 37:05 – Local Models: Mac Studio vs DGX Spark 39:03 – Reframing Cost as Investment 41:59 – The Real Way to Make Money With Hermes 42:51 – Closing Thoughts Key Points * Hermes Desktop pulls sessions, profiles, artifacts, skills, and cron jobs into one polished, Apple-style interface. * Smart session and context management keeps each message slim and keeps monthly costs low. * Profiles map to different models — Opus 4.8 for strategy, ChatGPT 5.5 for coding, a local Qwen model for free research — so each task runs on its best fit. * Reverse prompting plus a personal brain dump produces far stronger prompts, cron jobs, and outputs. * Sub-agents handle one skill across many parallel tasks; profiles handle work where each step needs a distinct skill set. * The biggest opportunity: aim your agent at Reddit and X to surface real problems you are positioned to solve. Numbered Section Summaries 1. The Moment Hermes Overtook OpenClaw Alex shares his screen and calls Hermes Desktop the best AI agent experience available right now, framing the recent release as the moment Hermes pulled ahead. He compares the focused, polished Hermes updates to Apple and the broader OpenClaw approach to Android, and notes his switch came purely for the love of the game. 2. Sessions and Context Management Alex shows how each new chat opens its own session, keeping context clean and separated by topic. He explains that slim context keeps every message small, and small messages keep monthly bills modest — his top money-saving tip. 3. Profiles Mapped to Models Profiles are separate Hermes agents, each with its own skills, a soul.md personality, and its own memories. Alex organizes his by model — Opus 4.8 for high-level strategy, ChatGPT 5.5 for coding, and a local Qwen model for free, fast research — and picks the profile whose strengths match the task. 4. Artifacts as a Productized Second Brain Artifacts gathers every link, image, and file you send your agent into one searchable place. Alex drops links to his "Librarian" profile and lets Artifacts file them automatically, which turns the agent into a tidy, searchable second brain. 5. Skills, Tools, and the Cron Section Alex walks the Skills interface, where 150-plus skills ship out of the box and you keep only the ones you use to trim context and cost. He highlights tool sets as a fresh way to group skills together, and praises the Cron section for giving one-click confirmation that scheduled tasks truly exist. 6. Reverse Prompting and the Brain Dump My favorite tip from Alex: brain dump everything about yourself — interests, goals, skills — then ask the agent for the best prompt to use. He demos it live by building a morning brief, and the reverse prompt returns a detailed, well-formatted instruction that pulls fresh, real headlines, prices, and scores from the last 24 hours. 7. Sub-Agents vs Profiles Alex clarifies the difference: sub-agents are copies of your main agent that share its skills, ideal for running one skill across many parallel tasks, like building several features of a micro-SaaS at once. Profiles each carry distinct skill sets and memories, ideal when a job needs a researcher, a writer, and a designer working in sequence. The #1 tool to find startup ideas/trends - https://www.ideabrowser.com/ LCA helps Fortune 500s and fast-growing startups build their future - from Warner Music to Fortnite to Dropbox. We turn 'what if' into reality with AI, apps, and next-gen products https://latecheckout.agency/ The Vibe Marketer - Resources for people into vibe marketing/marketing with AI: https://www.thevibemarketer.com/ FIND ME ON SOCIAL X/Twitter: https://twitter.com/gregisenberg Instagram: https://instagram.com/gregisenberg/ LinkedIn: https://www.linkedin.com/in/gisenberg/ FIND ALEX ON SOCIAL Youtube: https://www.youtube.com/@AlexFinnOfficial/videos X/Twitter: https://x.com/AlexFinnX Creator Buddy: https://www.creatorbuddy.io/

An LLM is an incredibly powerful brain that knows everything about humanity and nothing about you or the software you run. This video walks through the four buzzwords that turn that raw brain into a system you can trust: agent harness, loop engineering, LLM Ops, and eval. No jargon, step by step, technical or not. You can build anything. You can learn anything. 💪 Timeline 0:00 Intro: the 4 AI agent buzzwords 0:42 What an AI agent run actually is 1:52 The memory system: procedural, semantic, episodic 3:37 What "harness" really means (the horse) 5:07 Storing and updating memory (databases, skills, summarizer agent) 8:28 Retrieval: RAG, SQL vs semantic search 10:00 Tool calling and why agents loop 10:52 Loop engineering and end-loop guardrails 13:21 A Claude Code hooks example 14:18 Eval and LLM Ops: why you need them 15:55 Tracing every run (Langfuse, LangSmith) 16:43 Evaluation: LLM as a judge 17:20 Diagnosing what broke 18:16 The gate: ship the fix or fix the bug 18:57 Zoom out: the full system 19:53 Wrap up Launch Full-Stack App Template: https://github.com/ShenSeanChen/launch-mvp-stripe-nextjs-supabase My AI Agent Startup AutoManus.io is an AI sales lead manager for made-to-order product sales. We embed directly where conversations happen (WhatsApp, email, web chat) to capture inbound, automate follow‑ups, and kill CRM busywork; live with 4 pilots and 2 contracts in motion, pre‑seed backed by Character VC. Focused on US/UK/Germany, raising to expand engineering and scale deployments. ☕️ Buy me a coffee: https://buy.stripe.com/5kA176bA895ggog4gh - X: https://x.com/ShenSeanChen - Sean's Discord: https://discord.gg/tvECErKcFr - AutoManus Discord: https://discord.gg/5HhcNjCR - LinkedIn: https://linkedin.com/in/shen-sean-chen - Github Link: https://github.com/ShenSeanChen - Instagram: https://www.instagram.com/sean_ai_stories Sub Count when launched: 13,050 Top Github Repos - LaunchMVP NextJS,Supabase,Stripe (1015 stars): https://github.com/ShenSeanChen/launch-mvp-stripe-nextjs-supabase - ReactNextJSTips (114 stars): https://github.com/ShenSeanChen/yt-react-nextjs-tips - DeepResearchAgentBackend (70 stars): https://github.com/ShenSeanChen/yt-DeepResearch-Backend - SupabaseTipsKanban (64 stars): https://github.com/ShenSeanChen/yt-supabase-tips-kanban - RAGVectorDB (56 stars): https://github.com/ShenSeanChen/yt-rag - KimiK2Claude (37 stars): https://github.com/ShenSeanChen/yt-kimi-k2-claude - DeepResearchAgentFrontend (33 stars): https://github.com/ShenSeanChen/yt-DeepResearch-Frontend - AgenticRAG (13 stars): https://github.com/ShenSeanChen/yt-agentic-rag You Can Build Anything Playlist (Full-Stack Dev Without CS Degrees) - Core7: 95% React/NextJS Use Cases https://www.youtube.com/watch?v=vZzFlAjz4rA&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR&index=16 - Core6: 95% Supabase Use Cases https://www.youtube.com/watch?v=Pnoq2tIif3s&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR&index=15 - Core5: StripeNextJSSupabase https://www.youtube.com/watch?v=ad1BxZufer8&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR&index=8 - Core4: AuthSupabaseGoogle https://www.youtube.com/watch?v=AZAU3ClAWWE&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR&index=6 - Core3: FastapiGCPPro https://www.youtube.com/watch?v=u3sfqpzzo-U&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR&index=5 - Core2: SupabaseHero https://www.youtube.com/watch?v=V0kmgkyPsVQ&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR&index=4 - Core1: LaunchAnApp https://www.youtube.com/watch?v=iQ2izvGG4tQ&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR&index=3 - Launch WhereToMeet w/ GoogleMaps, Calendar, Supabase, Cursor: https://www.youtube.com/watch?v=5xMjX1QxGUY&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR&index=2 - Launch EncourageMe w/ Cursor, Supabase, Vercel https://www.youtube.com/watch?v=hnvL4xa2DHA&list=PLE9hy4A7ZTmpGq7GHf5tgGFWh2277AeDR

Get started with Codex, OpenAI’s coding agent, in this step-by-step onboarding walkthrough. You’ll learn how to install Codex, set up the CLI and VS Code extension, configure your workflow, and use Agents.md and prompting patterns to write, review, and reason across a real codebase. This video covers: 00:00 | Installing Codex (CLI + IDE) 04:00 | Setting up a repo and getting your first runs working 08:12 | Writing a great Agents.md (patterns + best practices) 14:14 | Configuring Codex for your environment 16:26 | Prompting patterns for more consistent results 27:02 | Tips for using Codex in the CLI and IDE 34:59 | MCP Configuration 46:50 | Advanced workflows: headless mode + SDK 52:02 | Codex resources to get you started Sign up: https://openai.com/codex/ Codex overview + docs: https://developers.openai.com/codex Codex Cookbook: https://cookbook.openai.com/topic/codex Install + setup VS Code extension: https://marketplace.visualstudio.com/items?itemName=openai.chatgpt Agents.md standard: https://agents.md Agents.md repo: https://github.com/agentsmd/agents.md Prompting + workflows Prompting guide: https://developers.openai.com/codex/prompting/ Exec plans (Agents.md patterns): https://cookbook.openai.com/articles/codex_exec_plans Config reference: https://github.com/openai/codex/blob/main/docs/config.md#config-reference Updates Changelog: https://developers.openai.com/codex/changelog Releases: https://github.com/openai/codex/releases

Hermes Agent is the best AI tool out there. Here is your COMPLETE tutorial. FULL Hermes Agent bootcamp in the Vibe Coding Academy coming up: https://www.skool.com/vibe-coding-academy 2nd Youtube Channel: https://youtube.com/@AlexFinnLabsOfficial Sign up for my free newsletter: https://www.alexfinn.ai/subscribe Follow my X: https://x.com/AlexFinn Henry Intelligent Machines (my new startup): https://meethenry.ai My $300k/yr AI app: https://www.creatorbuddy.io/ Prompts: I want you to schedule a task for yourself to do every single night at 2am. It should be a micro app, UI, or system that helps us get closer to my personal goals and ambitions. It should save us time, make us more productive, or just be generally useful. Make sure this is scheduled for every single night at 2am. hermes-agent.nousresearch.com Timestamps: 0:00 Intro 1:50 What Hermes Agent is 3:13 Hermes vs OpenClaw 5:17 Hermes vs Claude Code 7:57 Hermes installation 9:28 Choosing a model 13:44 First things to do 18:02 Hermes dashboard 21:00 Hermes Kanban board 23:15 Use case 1 25:40 Use case 2 27:47 Use case 3 29:40 Memory and skills 34:30 Mission control 37:38 Troubleshooting 38:42 Security