Following the completion of our very successful first session a couple of months ago, I am partnering with Packt Publishing Limited to run Cohort 2 of my live masterclass, Hands-On Harness Engineering with Claude, on Saturday, October 17, 2026, from 9:00 AM to 11:30 AM EDT (6:00 AM – 8:30 AM PDT / 1:00 PM – 3:30 PM UTC).
Frontier coding models generate code rapidly but fail in production through context decay, infinite retry loops, and unverified file modifications unless developers constrain them with external execution controls. I designed this curriculum around ten executable Python modules and an autonomous deep research capstone that pass 14 automated test suites and my five-gate production readiness scorecard. You can register through Packt’s Cohort 2 Eventbrite registration page with discount code kn40, providing 40% off attendance.
Packt and I deliver this 2.5-hour workshop as a live, interactive virtual lab with no local installations required to participate. Every registered attendee receives four core deliverables: (1) live instructor-led build, break, and repair exercises that I facilitate directly; (2) full access to the session recording after the event; (3) the complete open-source Python code repository containing all ten module implementations; and (4) an official Packt Certificate of Completion. You can also clone and run the exercises locally from the packt-harness repository on GitHub. Paid subscribers can get the 47-slide interactive presentation deck for free, with direct access links to both the slides and the GitHub repository provided in the paid section below.
In my work as Adjunct Professor at the University of San Francisco and CEO of DistributedApps.ai, I lead the OWASP AI Verification and Security Standard (AIVSS) project, co-chair two Cloud Security Alliance (CSA) AI Safety Working Groups, and contribute to the OWASP Top 10 for LLM Applications. I authored two books that establish this engineering foundation: Harness Engineering and Graph Engineering for Agentic AI Systems.
In this masterclass, I guide you through nine core technical modules:
Traditional Software Engineering Contrasted with Harness Engineering: the architectural shift from deterministic algorithms to probabilistic agent control environments.
The 5 Core Harness Pillars: memory tier precedence across five scopes and the 31-event PascalCase lifecycle map in Claude Code.
Spec-Driven Development: machine-readable SPEC.md contracts, allowed file scopes, and AST non-goal filtering.
Guardrails and Permission Gateways: PreToolUse command denial, AST secret scanning, and the append-only approvals ledger.
Tests as the Reliability Layer: the test-driven agent feedback loop, subprocess isolation, and anti-regression suites.
Agent Extensions and MCP Architecture: SKILL.md layout, plugin bundles, and Model Context Protocol stdio and Streamable HTTP security.
Compound Multi-Agent Teams: planner, implementer, and reviewer roles isolated inside ephemeral git worktrees.
The Five-Step SOP and the 5-Gate Production Readiness Audit: deterministic pipeline execution and objective readiness benchmarking.
Capstone Architecture: the Autonomous Deep Research Agent synthesized across an eight-stage execution pipeline.
