nit-band shows your prompt's English phrasing on Claude Code

The mod was built around a personal goal of working in English, where day-to-day requests would be made in English rather than the author's usual language. The workflow it targets is the ordinary one: sending prompts many times a day in a language other than the one being learned.

Rather than memorizing textbook sentences, the design surfaces the phrasing on screen above the input box while the agent is running, and stores only the expressions the user actually saves into a personal phrasebook at ~/.nit/phrasebook.md. Two hooks drive it on the Claude Code side — prompt.submit when a prompt is sent, and ui.render (AbovePrompt) for the strip above the input. The stripped prompt is sent once to Haiku without awaiting the call, so a slow or failed translation never blocks the prompt from being sent, and an incrementing counter keeps an older result from overwriting a newer one when prompts are sent in quick succession. Prompts written in the learning language, slash commands, code, and very short prompts produce nothing.

The design leaves the correction of English prompts out of nit-band, since the author mostly writes prompts in another language; that job was handed to a separate self-made skill. Whether the mod is useful to a given user likely hinges on how often they prompt in a language they are still learning, and on whether the phrasing the mod produces matches what their actual teammates would say.