Blog

Research Articles

Workflows

Graph Engineering for AI Agents: How to Set It Up with Skills

Agent loops hide the decisions that matter: what runs, what waits, which evidence counts, and what happens after a failure. Graph engineering makes them inspectable, and this guide ships eight reusable Skill templates for the core topologies.

Context

Context is infrastructure, not a prompt

The teams getting reliable output from agents stopped writing clever prompts and started engineering durable context. Here is the shape that work takes.

Context

Your AGENTS.md Is Not Documentation. It Is an Execution Contract

AGENTS.md should define the global operating contract for agent work, while routing files, stage contracts, skills, and references carry the rest of the context architecture.

Workflows

Reliable Agents Need Gates: Context, Plan, Review, Verify

One long agent chat is not an engineering process. Reliable agent work needs a loop with explicit gates, artifacts, and evidence.

Engineering

AI Does Not Fix Weak Engineering Systems. It Amplifies Them.

AI coding adoption is not a tool rollout. It is an operating-model change.

Reliability

Passing Tests Is Not Enough: The SlopCodeBench Lesson

Agent-generated code can pass today and still erode tomorrow. Teams need systems that measure maintainability, semantic fit, architecture drift, and proof of review.

Workflows

Stop Repeating Prompts. Turn Them Into Skills.

The best AI coding teams will not share prompt snippets. They will share versioned operating procedures that agents can execute.