Set up cloud dev environment + document setup in AGENTS.md

[ernop]

What & why

Sets up the Cursor Cloud development environment for MultiImageClient and records the non-obvious learnings for future agents. No product code changes — this PR only adds a ## Cursor Cloud specific instructions section to AGENTS.md.

Environment setup performed (in the VM/snapshot)

• Installed the .NET 9 SDK (9.0.316) via the official dotnet-install.sh to /usr/share/dotnet with a symlink at /usr/bin/dotnet. Ubuntu 24.04 apt ships only .NET 8/10, and Microsoft's Noble feed omits 9.0, so the script install is required.
• Created a local MultiImageClient/settings.json (gitignored, not committed) with the two required fields (LogFilePath, ImageDownloadBaseFolder).
• Update script (runs on VM startup): dotnet restore MultiImageClient.sln.

Verified working

Step	Command	Result
Restore	dotnet restore MultiImageClient.sln	OK (advisory NU190x warnings only)
Build	dotnet build MultiImageClient.sln	Build succeeded, 0 errors
Test	dotnet test MultiImageClient.Tests/MultiImageClient.Tests.csproj	Passed 21 / 21
Run (web UI)	dotnet run --project MultiImageClient/MultiImageClient.csproj -- --ui --ui-no-open	Kestrel on 127.0.0.1:5960; /api/config + / return 200

Hello-world action

All image providers require paid third-party API keys, none of which are provisioned in the cloud VM, so every generator is correctly reported as unavailable and disabled in the UI. As an offline core action I demonstrated the composer's built-in local typo-fix tool: typed a prompt with 4 deliberate misspellings and the app corrected them (photograpgh→photograph, feild→field, brihgt→bright, daylihgt→daylight) with a change report. A full successful image generation needs at least one provider key added to settings.json (cheapest: OpenAIApiKey).

Services

Primary product = the MultiImageClient console app + its built-in --ui web UI (both net9.0, cross-platform). MultiImageClient.Web is net9.0-windows (secondary; builds on Linux but cannot run here). The Django gallery and Python tools/ are optional/experimental and out of scope.