General feedback after running this

Feel free to close - just some examples encountered in the field to triage.

Let me know if these more unstructured reports are appreciated or too much bother to handle so i know on whether to keep doing this or not.

• i created a feature branch for this upgrade feature/net10 then it created another feature branch automatically called feature/net10-2 it should probably prompt me or something.

• Assesment dashboard did not update when the assement.md was present

  

• it ran aspire update --self but this has an interactive component where you need to pick the stable or prerelease so it got stuck there.

• Projects/dependencies were not populated during running

• Some chat/ui issues when retargeting to net10

  subagent;

  

  main;

  

  more of these empty codefences accumulated throughout the session

• it had some troubles restoring to and from github packages, the feed is defined in nuget.config and should be able to restore when authenticated to GH. I am authenticated. it tried various GH commands including one that messed around with the token, this seemed scary but hey i ran it under autopilot so.

• EF core provider pomelo.sql was still on net9 which means 0 compat with net10 and efcore10 - the assistant should mention this. I recovered by instructing to change the package to a fork but i feel like. https://github.com/PomeloFoundation/Pomelo.EntityFrameworkCore.MySql/issues/2007 -> edit; i did not see this cause the ui did not load the assesments but it was mentioned in the initial assesments.md but the ai took no course of action to recover to prompt me on what to do.

• Some timing indication on how long this would typically run for small to large apps, you need to kinda babysit this thing.

• Token cost guidance/indication

• Guidance on model picking, i ran autopilot with sol but i'd hope the executor tasks would use tera for example, im not sure this happens. The readme has no guidance on what models to pick.

• it blocked vscode:// so i could not open the files directly. The links open from my browser so i'm not sure why this was blocked or did not work. Is it vscode integrated browser or?

• Im impressed by how this goes through a guided step committing in between and so on, a blog post/architecture info regarding how this works including the sub agent handoff would be cool to learn from and create your own workflows like these.

• It spend a long time adding transitive package System.Security.Cryptography.Xml and updating it to version .11 presumably cause the one that comes with the fx has a vulnerability. This should better be left as future work, i need the upgrade to work first.

• monorepo support; the executor agents seem to coninuously start in the repor root, fail to run commands and then cd into src\front-end (angular frontend + .net backend)

• Kept getting The terminal process "C:\Program Files\PowerShell\7\pwsh.exe" terminated with exit code: <...>. modal

• It generate 4-5 tasks of removing

  <PackageReference Include="Microsoft.Extensions.ApiDescription.Client">
  			<PrivateAssets>all</PrivateAssets>
  			<IncludeAssets>runtime; build; native; contentfiles; analyzers; buildtransitive</IncludeAssets>
  		</PackageReference>
  

  per project instead of a single step making it take a long turn.

• Under 04-build-and-fix-compatibility the tests steps are split (per project/feature) so it tests and fix each, but it looked like half the project already worked so it could skip a lot, perhaps a validate before this fan-out.

• The vs code native file-edi-tracker sees a lot of files and it makes the ide kinda slow, i clicked 'keep' now and then

• A cool adition would be to check github workflows/actions and make sure CI still works, or atleast point out possible issues.

• I have a sourcegenerator project, im not sure why but the process made it so it would lock this project. I taskkilled a .net host process. It could have been the embedded dashboard since that stopped working after, but it was also the final step i was on, it also recovered.

• i was on about 300k context window tokens, should it not /compact a bit inbetween?

• It was creating a lot of files like under .validation for example, could these not have a .gitignore file added in the folder so that it would not push 2k+ files into memory for the git changes view in the ide?

• an option to clean after success so it removes the artifacts under .github/upgrades

Start by splitting this unstructured report into separate issues, using the named areas such as assessment.md, nuget.config, README, .github/upgrades, and .validation as entry points. Reproduce each reported behavior before defining a focused acceptance condition; this report has no single completed state.