Harness Engineering: I completed the SULICOM workshop

I completed SULICOM’s Workshop - Harness Engineering. Certificate WS-Y5KHH2W47Z was issued on September 29, 2026.

A harness is the system around a model that makes it a working agent: the system prompt, tools, agent loop, context management, rules, skills, and hooks. The workshop approached this system from two sides.

What the program covered

• The harness from the inside. Participants built a minimal agent loop in TypeScript without frameworks, adding tools, message history, request tracing, rules, and skills. The program also covered prompt injection protection and a hook requiring approval before a file is written.
• The harness as a variable. The workshop compared how Claude Code, Codex, and other agents load instructions and use hooks, MCP, and skills. The practical work included an agent with memory, tool argument validation, and logging.

For me, the value of this topic is that agent quality depends on more than the model. What context does it receive, what actions can it take, and how is the result checked? I will test these practices in my own engineering work before drawing conclusions about their effectiveness.

Full program and certificate verification.