iii-hq/workers — reverse-engineered prompt
Reverse engineered prompt
Build me a modular worker toolkit for the iii engine. I want each top level module to act like its own worker process that connects over WebSocket, registers functions and triggers, and can be installed on its own from a workers registry. The goal is to give the engine a bunch of useful abilities out of the box, like shell and filesystem access, database queries, object storage, image resizing, web scraping, session and conversation history, context building, approval gating for human review, and chat delivery through Slack and Telegram.
I also want the AI side covered. Add a routing layer for model calls with provider support, and wrappers for agent style tools like Claude Code, Codex, Grok, Devin, OpenCode, Pi, and Hermes so their events can be streamed back into the system in a consistent way. It should also include developer quality of life pieces like an MCP bridge, language server support, and installable agent skills for Cursor or Claude Code. If anything is unclear, look up the current iii docs and match the existing repo structure and intent.
Want more depth? Deep Reverse