The Executable Harness
Some agent rules belong in scripts rather than prose — filename suffixes that encode permission, and linters that keep the harness from rotting.
Thinking out loud
Frameworks, workflows, and lessons from building and operating products.
The Agentic Harness in Practice
A walkthrough of one solo-operator codebase mapped onto the four-level harness model — the artifacts, scripts, and negative-space rules that make each level real.
Handling Context Compaction
What coding agents lose to compaction, resumes, and long goal-driven loops — and the disk-anchored protocol that keeps work correct across them.
Multi-Agent Workflow Patterns
Four structured workflows — research, implementation, documentation, and security — that use multi-agent pipelines with built-in review cycles.
The Agentic Project Harness
A baseline operating layer for agent-friendly repositories — four readiness levels from context harness to product harness.