UPstream fetch daily maintaines

Merged
A-
2d ago
+1701-649

Issue for this PR

Closes #

Type of change

  • Bug fix
  • New feature
  • Refactor / code improvement
  • Documentation

What does this PR do?

Please provide a description of the issue, the changes you made to fix it, and why they work. It is expected that you understand why your changes work and if you do not understand why at least say as much so a maintainer knows how much to value the PR.

If you paste a large clearly AI generated description here your PR may be IGNORED or CLOSED!

How did you verify your code works?

Visual

Optional, for user-facing changes: a small looping animated SVG (240×72) of the change, shown in the desktop app's "What's new in beta". Replace the block below or delete this section.

Screenshots / recordings

If this is a UI change, please include a screenshot or recording.

Checklist

  • I have tested my changes locally
  • I have not included unrelated changes in this PR

If you do not follow this template your PR will be automatically rejected.

Summary by cubic

Bundles several independent fixes and refactors across the AI loop, skill handling, the Claude Agent provider, the MCP layer, and the CLI updater.

  • Refactors the AI tool loop: malformed tool inputs go back to the model as errors for retry, token-limited steps end without running their tool calls, and AIError gains status, code, isRetryable, and retryAfterMs.
  • Moves built-in and virtual skills to skill:// URIs so they read as in-memory files in the read tool, link resolver, file paths, and open-in-app.
  • Fixes compaction so a rejected conversation is not resent unchanged, and disables Claude Agent's non-streaming and refusal fallbacks while reporting plan usage on step metadata.
  • Loads skills and instructions from a connected OpenCode Console, redirects the Go, Zen, pricing, models, and teams pages to Console pages, and adds the Console sitemap.
  • Splits MCP connection lifecycle from its location handling.
  • Adds --allow-scripts to npm global installs so post-install scripts run on upgrade.

Summary by CodeRabbit

  • New Features
    • Connected Console workspaces can now provide shared instructions and skills.
    • AI error details now include status, provider code, retryability, and retry delay.
  • Improvements
    • AI tool loops now handle invalid tool arguments and execution failures as error results, respect output limits, and report usage across steps.
    • Skill files can be read and displayed through skill:// paths.
    • Console pages now redirect to their updated destinations.
  • Bug Fixes
    • Context-overflow recovery can continue with a text-based summary request.