Agentic Harness Lab: Screen Recording 2026-08-27 8:04 AM

This screen recording is part of the Agentic Harness Lab, a comparative engineering project supporting my Drupal GovCon 2026 session, “Anatomy of an Agentic Harness: Building the Same Agent in Drupal AI, LangGraph, and CrewAI.” The project builds the same Drupal alt-text review workflow in three different agentic frameworks and compares how each one handles six parts of an agentic harness: Context · Tools · State and Memory · Verification · Human Review · Lifecycle and Recovery These recordings document the work as it happened, including implementation, testing, debugging, package execution, evidence collection, and the engineering decisions made along the way. They are working-session records rather than polished tutorials, so some videos may include repeated commands, intermediate states, troubleshooting, and incomplete work. Project repository: https://github.com/dhouse109/agentic-... The repository contains the source code, experiment specification, implementation plans, shared contracts, evidence, claims register, and comparison materials used for the project. This project follows an evidence-first approach: expected framework behavior begins as a hypothesis, and conclusions are updated based on what the pinned implementations actually demonstrate. #AgenticAI #Drupal #DrupalAI #LangGraph #CrewAI #AIAgents #HumanInTheLoop