ndom91/open-plan-annotator — reverse-engineered prompt
Reverse engineered prompt
Build me a local first plugin for coding agents that lets me review an agent’s plan before it starts making changes.
When the agent finishes a plan, it should open a browser window with the plan in a clean review UI. I want to be able to select text and mark it for deletion, replacement, insertion, or leave a comment. Then I can either approve the plan so the agent continues, or request changes so the annotations get sent back as structured feedback and the agent can revise the plan. This should work in a loop until I approve it.
Please make it run entirely locally, with a small temporary server and a web UI. It should support Claude Code, OpenCode, and Pi integrations, plus a manual command where I can review a markdown plan file directly. Include basic install commands, a doctor command, keyboard shortcuts, and enough docs so someone can use and maintain it. Use TypeScript, and look up current plugin docs online if needed.
Want more depth? Deep Reverse