[Feature]: Audit non-terminal Windows OS capability gaps from #2983

[neubig]

Why

Some Windows-related differences exposed by #2983 are not really terminal problems at all. They come from underlying OS capability differences such as symlinks, POSIX permissions, fork(), or resource semantics.

Those cases should be tracked separately from terminal support so we can decide which ones need:

• a portable rewrite,
• a Windows-specific assertion, or
• an explicit capability marker.

Scope

Examples touched by #2983 include:

• symlink-dependent tests
  • tests/tools/file_editor/test_basic_operations.py
  • tests/sdk/skills/test_skill_utils.py
• POSIX permission semantics
  • tests/tools/file_editor/utils/test_file_cache.py
• fork() / resource / process-isolation assumptions
  • tests/tools/file_editor/test_memory_usage.py
  • tests/sdk/conversation/local/test_conversation_core.py

Goal

Audit these cases one by one and decide whether they should be:

1. rewritten to be portable,
2. asserted differently on Windows, or
3. represented by a shared capability marker/helper instead of scattered ad hoc skips.

Suggested approach

• inventory each Windows guard added in Add Windows test support #2983 that is unrelated to the terminal runtime
• separate “environment not configured” from “OS does not support this concept the same way”
• centralize any remaining capability checks to keep the test suite easier to audit

Related

• PR: Add Windows test support #2983
• terminal runtime issue: [Feature]: Implement Windows terminal backend and PowerShell-aware behavior #2987
• terminal skip cleanup issue: [Feature]: Remove terminal-related Windows skipifs added in #2983 #2986

This issue was created by an AI agent (OpenHands) on behalf of the user.