[codex] fix sandbox glob_files recursive patterns

[gaoxiaolei-s59]

Summary

• normalize leading **/ in sandbox glob() before passing the pattern to find -name
• add a regression test covering glob_files calls like **/*.md

Root Cause

The glob_files tool description advertises recursive patterns such as **/*.java, but BaseSandboxFilesystem.glob() forwarded that pattern directly to find -name. find -name matches the filename portion only, so the leading **/ caused sandbox-backed recursive glob calls to return no matches.

Validation

• mvn -pl agentscope-harness -am -Dtest=BaseSandboxFilesystemTest,FilesystemGlobTest test

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

❌ Patch coverage is 50.00000% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...gent/filesystem/sandbox/BaseSandboxFilesystem.java	50.00%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!

[AgentScopeJavaBot]

🤖 AI Review

This PR fixes a real bug: glob_files in sandbox mode advertised support for recursive patterns like **/*.java, but the implementation forwarded the pattern directly to find -name, which only matches the filename portion. The fix correctly strips the leading **/ prefix since find already performs recursive directory traversal. The change is minimal, null-safe, and well-tested with a new BaseSandboxFilesystemTest using a clean fake-sandbox design. One minor symmetry observation regarding the sibling grep method is noted below.

[AgentScopeJavaBot]

[nit] Symmetry note: the sibling grep() method (line ~282) passes its glob parameter directly to grep --include= without the same **/ stripping. GNU grep's --include also matches only the filename portion, so grep_files with **/*.java would silently return no matches — the same bug this PR fixes for glob_files. Consider applying the same normalization in a follow-up, or extracting a shared helper like stripRecursivePrefix(String).

[AgentScopeJavaBot]

[praise] Nice test design — capturing the generated command string and asserting it directly is the right approach for testing shell-command delegation. The FakeSandboxFilesystem stub is minimal and focused.

[AgentScopeJavaBot]

🤖 AI Review

This PR fixes a real bug: glob_files in sandbox mode advertised support for recursive patterns like **/*.java, but the implementation forwarded the pattern directly to find -name, which only matches the filename portion. The fix correctly strips the leading **/ prefix since find already performs recursive directory traversal. The change is minimal, null-safe, and well-tested with a new BaseSandboxFilesystemTest using a clean fake-sandbox design. One minor symmetry observation regarding the sibling grep method is noted below.

[AgentScopeJavaBot]

[nit] Symmetry note: the sibling grep() method (line ~282) passes its glob parameter directly to grep --include= without the same **/ stripping. GNU grep's --include also matches only the filename portion, so grep_files with **/*.java would silently return no matches — the same bug this PR fixes for glob_files. Consider applying the same normalization in a follow-up, or extracting a shared helper like stripRecursivePrefix(String).

[AgentScopeJavaBot]

[praise] Nice test design — capturing the generated command string and asserting it directly is the right approach for testing shell-command delegation. The FakeSandboxFilesystem stub is minimal and focused.