GammaLabTechnologies/harmonist — reverse-engineered prompt

Build me a portable AI agent orchestration kit for coding assistants like Cursor, Claude Code, Copilot, Windsurf, and Aider.

I want it to live inside a project with no runtime services, no database, and no extra dependencies beyond Python standard library and shell scripts. The main idea is that AI agents can’t just say they followed the rules. Hooks should actually check that required reviewers ran, QA happened, memory was updated, and edited files are still trusted. If something is missing, the assistant should get a clear follow up message and the task should not finish yet.

Include a catalogue of specialist agents that can be selected by role or domain, plus a simple integration prompt that lets an AI assistant install and configure the right agents for a project. Add structured memory with validation and secret scanning, supply chain verification using a manifest, cross platform hook support, docs, examples, and tests. Keep it simple, auditable, and easy to drop into an existing repo.