Claude for Leaders: Accelerate Your Business

Understand what GitHub is and what version control actually does, then bring your own knowledge base and building blocks — skills, agents, prompts, markdown — under it. You'll read a real repository and follow a change through it end to end before creating anything yourself, so the commands you run afterwards make sense rather than being copied.

Master the anatomy, capabilities, and basic orchestration patterns of AI agents so you can reason about any agent system. Then translate that foundation into practice by analyzing Claude Code's subagent architecture and building your first working subagent with the right tools, instructions, and permissions. By session's end you'll have a foundation for thinking about agents and a running subagent on your own machine.

Build a foundation for thinking about agents — and your first running one. You'll analyze the anatomy that shows up in every AI agent, map it onto Claude Code's subagent architecture, and build your first Claude custom agent, in Claude or Claude Code, for a use case you care about. By the end, you'll have a durable mental model for reasoning about any agent system and an agent running on real work.

Automate repetitive web tasks end-to-end. You'll build browser automation workflows using Claude in Chrome for data extraction, form filling, and multi-step processes, then implement scheduled and event-driven triggers that run autonomously—creating workflows that execute on their own without manual initiation, freeing you from repetitive browser-based work completely.

Compose multi-agent systems in Claude Code from a Building Block Spec, choosing sequential, parallel, or hybrid execution patterns to fit the work. Run the system in augmented mode (with you steering at checkpoints) or automated mode (unattended), and operationalize automated runs with native Claude Code / Desktop scheduling. Then close the loop by analyzing execution traces to diagnose failures and refine subagent configurations.

Build a working multi-agent system in Claude Code. You'll compose multiple subagents from a workflow spec, implement orchestration patterns that fit the work, schedule automated runs, and debug execution from traces. By the end, you'll have a multi-agent system you can run on demand or unattended — without venturing into full agent teams.

Give Claude more than memory — it has that now. Give it your business, organized: your information structured and linked so Claude can reason across it. Until now every conversation started from zero, the judgment you built lost when the tab closed. Learn the compounding knowledge graph pattern — an LLM wiki in Google's Open Knowledge Format — then build your own from an empty folder: Claude interviews you, proposes concept types, builds the graph. Then run the loop live: query, ingest, lint.

Stop starting every AI conversation from zero. You'll build a second brain for your business — linked markdown pages in Google's Open Knowledge Format that Claude writes, cross-links and keeps current for you — following the LLM wiki pattern Andrej Karpathy published. Prove it works with a cited query answered from your own pages, and leave with the foundation every workflow you build reads from and writes to.

Stand up your plugin distribution pipeline end-to-end. Set up a GitHub repo for your skills and agents, package them as an installable plugin with Plugin Dev's guided workflow, and ship the plugin into Cowork so your team can install it with one command. By module end, your plugin is distributed and any change you push reaches your team's Cowork workspaces without manual upload.

Build a personal knowledge base that compounds value with every use, implement a learning-loop pattern so Claude gets smarter session over session, then package your skills and agents as an installable plugin and ship it across Cowork and Claude Code so your AI follows you everywhere — without manual file uploads.

Run the Business-First AI Framework end-to-end yourself, live in Cowork. Starting from a pre-built workflow definition, invoke /design, /build, /test, and /run on a Competitive Intelligence workflow — shipping a competitor-research skill and competitor-brief agent that produce a brief on a sample competitor. By session end, you've run every framework step hands-on and shipped a workflow ready to evolve into a self-improving system next session.

SRun the Business-First AI Framework end-to-end yourself, in Cowork — taking a Competitive Intelligence workflow from a pre-built definition through Design → Build → Test → Run via the framework's slash commands. Generate a competitor-research skill and competitor-brief agent from your approved spec, run them, and watch a structured context file emerge — the foundation for next session's self-improvement work. Close with a brief preview of Step 7 (Improve).

Take the workflow you shipped in Session 9 and evolve it into a self-improving system. Apply the LLM Wiki pattern to your workflow's context file — adding Rules / Facts / Hypotheses sections, an ingest/query/lint loop, and a daily schedule so it keeps improving on its own. This is Step 7 (Improve) in action. Then reflect: share what you've applied, hear what landed for peers, complete the survey, and chart your next steps as an AI builder.

Take the Competitive Intelligence workflow you shipped in Session 9 and evolve it into a system that gets sharper every time it runs. Apply Karpathy's "LLM Wiki" pattern to its structured context file — structure it with Rules / Facts / Hypotheses sections, implement an ingest/query/lint learning loop, and schedule the workflow to run daily without you. By module end, your workflow's outputs feed a self-improving knowledge base of competitor intelligence — Step 7 (Improve) in action.

Trace the cumulative builder stack you've assembled across five weeks, then share what you've actually applied to your work and hear what landed for the rest of the cohort. Complete the survey and chart your next steps — leaving with a clear picture of how far you've come and where you're going next.