Upload a local file

Use this when the user asks to upload a file from their own computer to SimplyPrint (STL, 3MF, gcode, etc.). This uses a small helper script because MCP tools cannot read files off the local disk.

The helper targets files.simplyprint.io — the API-facing file-upload service. It returns a hex hash (the API file id) which can later be passed to queue/AddItem as fileId or printers/actions/CreateJob as file_id.

Requirements:

• SIMPLYPRINT_API_KEY and SIMPLYPRINT_COMPANY_ID must be set in the environment. If they are missing, stop and tell the user to set them (point at docs/env-vars.md in the plugin folder).
• The helper script lives at ${CLAUDE_PLUGIN_ROOT}/scripts/upload.sh on macOS/Linux/Git Bash, and ${CLAUDE_PLUGIN_ROOT}/scripts/upload.ps1 on Windows PowerShell 7+.
• The user's account must be on the Print Farm plan (files API requirement).

Steps:

1. Confirm the local file path the user wants to upload. If they only gave a file name, ask.
2. Pick the helper: on Windows (non-Git-Bash shells), use the .ps1; otherwise use the .sh.
3. Run the helper with --file <path> (or -File on PowerShell).
4. Parse the single stdout line FILE_ID=<hex-hash>. Everything on stderr is human-readable context, not structured output.
5. Confirm back: file name, file id (hex hash).
6. Ask the user what they want to do next — queue it (queue-print with --file-id), start it on a printer (start-print with --file-id), or just keep the id for later.

If the helper exits non-zero, show the stderr text to the user verbatim. Common causes: missing env vars, wrong company id, revoked API key, file too large (>100 MB per part — the helper currently uploads single-part only), account not on Print Farm plan.

Do not try to reimplement the upload by calling raw HTTP yourself. Always go through the helper script so behaviour stays consistent with the other flows (start-print, queue-print).