Technology notes

The core playbook is stack-neutral. These files hold what we learned about specific tools. Read a file only when your stack uses that tool. Each one opens by saying when it applies, then gives the general pattern and the setting we used on the reference project.

File Read it when your stack uses
claude-code-harness.md Claude Code as the orchestrator, with subagents (always, for this playbook)
windows-powershell.md A Windows workstation, PowerShell 7 scripts, Git Bash beside them
dotnet.md .NET SDK builds and tests (MSBuild, Microsoft.Testing.Platform, xUnit)
postgresql.md PostgreSQL, Liquibase or other changelog migrations, schema-per-module ownership, plan tests
docker-and-local-stack.md Docker Desktop, Compose dev stacks, Testcontainers, a secrets agent, local TLS
playwright-e2e.md Playwright end-to-end tests against a seeded local stack
gcp-demo.md A shared demo hosted on a single cloud VM (we used a Google Cloud free trial)

Figures and incidents from the reference project appear only as evidence. You don’t need to know the project.


Table of contents


Back to top

MIT licensed. Distilled from a real AI-orchestrated build.