louislva/prompt: The Surgical Context Injector for Terminal Purists

The Alt-Tab Tax

The modern developer workflow is fracturing under the weight of context switching. Jumping between a terminal, an editor, and a browser tab to converse with an LLM incurs a heavy cognitive toll. This friction is primarily driven by the need to manually copy and paste hundreds of lines of code to provide the AI with adequate context.

While the industry races toward ubiquitous "Agentic IDEs" that index every file on a hard drive, `louislva/prompt` takes a radically different approach. It argues for the surgical strike. By bringing `@-mentions` directly to the raw terminal, it solves copy-paste fatigue without the bloat of a heavy editor.

Scoring by the Clock: The Recency Heuristic

The most compelling technical detail of this 10KB script is its sorting algorithm. When a user types `@` in the terminal, the tool does not rely on complex vector search or semantic relevance. Instead, it relies on a much higher-signal metric for active coding: recency.

Using `os.path.getmtime`, the script surfaces the files modified most recently. Developers usually want to prompt the AI about the files they just changed, not a random dependency buried deep in a monolithic folder structure. This acknowledges that a developer's short-term memory is the most valuable context for an LLM.

Inside the 10KB Engine

The architecture is aggressively minimalist. The core engine resides in a single `main.py` file. It leverages `prompt_toolkit` to create an interactive shell and uses a custom `FilePathCompleter` to handle the `@` logic.

The parsing logic uses regex lookbehinds to allow literal `@` symbols while still triggering file injection when intended. Once the prompt is assembled, the tool reads the referenced files, formats them into Markdown code blocks, and pipes the entire payload directly to the system clipboard.

The Invisible Hand: Style Prompts

Beyond simple file injection, the tool introduces a persistent configuration layer via `user_settings.json`. This acts as a lightweight system prompt for the terminal environment.

By defining a global `style_prompt`, developers can automatically append specific personas or formatting rules to every query. It provides the power of custom instructions for developers who prefer to live entirely within a multiplexer like `tmux`.

Minimalism as a Power Feature

In a landscape dominated by 500MB Electron applications, a tool with a sub-megabyte footprint stands out. The reliance on standard Python libraries and a strict adherence to doing one thing well makes it incredibly resilient.

For specific, high-speed prompting tasks, the invisible tool often outperforms the omnipresent one. It proves that sometimes the best way to interact with cutting-edge AI is through the oldest interface we have.