My engineering workflow
How I build.
Project context, reusable skills, and AI agents. A look at the process behind the work, and the judgment that connects it.
Agents help with the work.
I own the decisions.
01 / The process
From intent to delivery.
Select a stage to explore.
An illustrative walkthrough, not a live run.
Stage 01 / 06
Start with the right problem.
Good agent work starts before the first prompt. Define the outcome, read the project context, and make the boundaries clear.
- The input
- A business need, a bug, or a feature request.
- Agent contribution
- Read the repository, gather context, and surface missing requirements.
- My responsibility
- Choose the outcome, constraints, and acceptance criteria.
Add a workflow page to the portfolio. Reuse the design system, preserve existing assistant routes, and support keyboard and mobile use.
Project docs · AGENTS.md
02 / The toolkit
A tool for each kind of work.
Start with the right skills.
Reusable instructions bring specialist guidance into a task. Pick the relevant skill, read its rules, and apply it within the project’s context.
Explore skills.sh (opens in a new tab)Design guidance · Implementation patterns · Review checks
- T3 CodeWorkspace for coding with AI agents.Coding workspace (opens in a new tab)
- OrcaWorkspace for coding agents and parallel worktrees.Agents & worktrees (opens in a new tab)
- OpenBotDesktop workspace for a team of AI agents.Local agent teams (opens in a new tab)
- Grok BotAI agent for tasks across apps and websites.Tasks across apps (opens in a new tab)
03 / My skills
Specialist guidance, on demand.
These are the skills in my toolkit. I choose them to fit the task: from shaping a brief to reviewing the final change.
41 skills / 8 groups
04 / A closer look
Inspect the assistant behind this site.
Read how the portfolio assistant is built and evaluated, then explore the steps observed during a real question.