FDK File Fetcher

Architecture

Target flow:

1. User gives input (file path, extension, call path, mode, optional output).
2. Agent runs the script with those inputs.
3. Script fetches the file from GitHub only (no local fallback).
4. Script writes the result to stdout (chat: source or JSON; create: JSON).
5. Agent reads stdout and returns the fetched response to the user.

All source is always fetched from GitHub; the script does not read from local files or node_modules for content.

Requirements