What you'll get out of this workshop
The workshop is built around three things working developers actually need from AI coding tools:
- Get better answers from AI coding agents. Package the files, traces, screenshots, logs, and goals that help AI tools understand the real problem.
- Master terminal-based agent workflows. Use terminal sessions for research, planning, building, and validation without losing the thread.
- Create specialized agents for real work. Shape prompts, skills, and tool access so each agent has a clear role instead of doing everything.
What people say
John's workshop revealed that I was barely scratching the surface. I'd used features like agents and hooks before, but never to this extent. It was top notch! Insightful, straight to the point and brilliantly structured.
Vitor Correa
This workshop was a goldmine of practical tips and tools that will take my workflow to the next level.
Jan
JOHN IS SO SMART!
Kent C. Dodds
Curriculum
A practical map of the whole agentic stack.
Context: get better answers from hard problems
Package the right project context, ask for ambitious plans, and turn the response into an implementation-ready goal.
- Bundle files, logs, issues, screenshots, and traces
- Ask for plans that are specific without being timid
- Use the best plan as the source for downstream work
Workspace: terminal workflows for parallel agents
Use named terminal sessions for researchers, planners, builders, and validators.
- Create a primary terminal that coordinates focused agents
- Fan work out to research and build lanes
- Keep parallel work easy to scan and compare
Profiles: specialized agents beat overloaded agents
Compare broad default agents against narrow custom prompts, focused skills, and limited tool access.
- Assign planner, builder, researcher, and validator responsibilities
- Control instructions, skills, and available tools
- Measure cost, speed, and result quality tradeoffs
Tools: connect your tools into one AI development system
Organize local tools, remote services, and command-line utilities so each agent gets the right level of access.
- Organize tools by agent type and risk level
- Separate read-only tools from mutation tools
- Map agent configuration to repeatable tool profiles
Exploration: explore, debug, and decide with richer context
Replace vague back-and-forth with fast variations, browser context, screenshots, recordings, and selective pruning.
- Generate multiple directions before deciding
- Point agents at the exact UI state you want investigated
- Turn visual options and rough recordings into concrete goals
Extensibility: reusable skills, hooks, and workflow automation
Create portable expertise, package repeatable workflows, and know when a flexible agent process should become a script.
- Create, share, install, and manage reusable skills
- Use hooks and validators to enforce important behavior
- Turn repeatable agent work into lightweight automation
Memory: capture goals, decisions, and history
Connect written goals, task trackers, and past conversations so future work starts with the right context.
- Decide when lightweight local goals are enough
- Move durable work into a team-friendly tracker
- Mine past conversations for reusable decisions
SDK: build Codex into your own tools
Use the Codex SDK when you want programmatic control over engineering agents inside internal tools, apps, and delivery pipelines.
- Start and continue Codex threads from TypeScript applications
- Integrate agent runs into CI, dashboards, and team workflows
- Know when the SDK is better than one-off terminal sessions
Outcomes
Leave with an operating system, not a notes doc. The workshop is built around concrete workflows you can reuse: context packaging, specialized agents, tool profiles, hooks, goals, browser context, and repeatable scripts.
- A repeatable context-packaging workflow for hard problems
- A terminal workflow pattern for parallel specialized agents
- A mental model for planner, builder, researcher, and validator roles
- A tool configuration strategy for local commands, external services, and AI agents
- A practical approach to reusable skills, hooks, goals, and conversation history
- A starting point for embedding Codex into internal tools with the SDK
