Agentic Engineering Workshop
Master AI coding agents. Three days that change how your team builds software, from first prompt to production PR.
Small groups, serious practice.
No watching a slide deck for six hours. Every session sets up a lab; every lab gets into real code with a real agent.
- Claude Code · CLI
- Codex CLI · CLI
- GitHub Copilot · CLI
- Gemini CLI · CLI
- Junie CLI · CLI
- OpenCode · CLI
- Cursor · Orchestrator
- Conductor · Orchestrator
- Superset · Orchestrator
- Antigravity · Orchestrator
- JetBrains Air · Orchestrator
- Agent HQ · Orchestrator
Six durable skills, not six vendor tricks.
The tooling will change. These won't. Every skill is taught against your codebase, in your agent, with a concrete rubric for “done.”
Read Agent Behavior
Detect drift, hallucination, and context decay before they compound into trust debt. Map capability ceilings before you delegate.
Write Effective Plans
Create specifications for short-horizon and long-horizon tasks, with constraints, checkpoints, and definitions of done.
Validate Systematically
Apply a six-layer validation framework that goes beyond “it compiles and tests pass”, including automated trust thresholds.
Intervene at the Right Moment
Know when to let the agent run, when to redirect, and when to stop. The skill that prevents wasted cycles.
Manage Agent Fleets
Decompose work across multiple agents using Git worktrees. Spawn, supervise, and reap. Fleet management, not one-shot tools.
Build Corporate-Ready Workflows
Agent instruction files as team infrastructure (works for Claude Code and GitHub Copilot), MCP integration, agent credential security, scheduled and event-triggered automation, and the autonomy progression.
One question, three acts.
Agents get you to ~70% correct. The workshop is about the other 30%: three days, each a durable skill that closes the gap.
- TALK The 70% problem & the three durable skills
- LAB Lab 1 · Watch the Machine (observed)
- TALK How agents actually work: the context model
- LAB Lab 2 · Context archaeology
- TALK The tools: CLIs, orchestrators, MCP
- LAB Lab 3 · Write a plan an agent can execute
- TALK Working with plan mode
- TALK Context engineering: layers, budgets, resets
- LAB Lab 4 · The measured build
- TALK Intervention and control: nudge / redirect / reset
- LAB Lab 5 · The drift (detect and recover)
- LAB Lab 6 · The constraint (corporate standards)
- TALK Validation: the six-layer checklist
- TALK Parallel work & multi-agent patterns
- LAB Lab 7 · Parallel agents with worktrees
- LAB Lab 8 · The recovery (a failing session)
- LAB Lab 9 · The capstone: synthesize
- TALK The corporate layer & honest cost-benefit
Two ways to run it. One clear price.
A private workshop delivered for your team on-site or remote, or pool interest for the next public workshop.
Per engineer
- 3 days · 1:2 theory-to-lab ratio
- Committable AGENTS.md entry that works in Claude Code, Copilot, Codex, Gemini, Junie, and OpenCode
- Plan Template (with worked example) + six-layer validation checklist
- Tool Map across agentic CLIs and orchestrators
- First-30-days practice guide, week by week
- Held in Zurich, Bern, Basel, Chur, or St. Gallen · venue follows the group
- Catered lunch and breaks (on-site only)
6-7 engineers · on-site or remote
- Everything in the per-engineer package
- Pre-workshop briefing with your team lead to align examples and constraints with your environment
- Corporate-readiness + regulated-industry walkthrough calibrated to your stack
- 30-day follow-up review with your team lead
Trust debt, paid down.
AI coding agents promise velocity. Without the right skills, they deliver trust debt: code that looks correct but fails under pressure, accumulating invisible cost that your most senior engineers pay later.
This workshop teaches the supervision skills that make agent-assisted development sustainable: reading agent behavior, writing plans agents can follow, validating output systematically, and building the verification infrastructure that lets your team scale with confidence.
Tell us about your team. We'll reply within 48h.
Tell us the shape of your team, the stack you work in, and when you'd like to run it. We'll reply with dates, logistics, and an honest read on fit.