IgorWarzocha/Opencode-Context-Analysis-Plugin ? reverse-engineered prompt

Build me an OpenCode plugin that helps me understand where my AI tokens are going during a session.

I want a /context command that gives a clear breakdown of token usage for system prompts, my messages, assistant replies, tool outputs, and reasoning traces, with simple charts or a readable summary so I can spot what is using the most context. It should also show which tools like read, bash, or webfetch are costing the most, and let me ask for a quick summary or a more detailed view.

Please make it work locally, without sending anything to external services, and keep it easy to install by copying the plugin into an OpenCode project or a global OpenCode config. If you need to check the latest OpenCode plugin docs or tokenizer docs online, go ahead.