fix(opencode): reject path traversal in remote skill discovery

[1837620622]

Issue for this PR

Closes #36365

Type of change

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

What does this PR do?

Remote skill indexes could supply skill.name / files with .., absolute paths, or cross-origin URLs. packages/opencode wrote downloads via path.join without containment checks.

Port the path safety checks already present in packages/core skill discovery (isSafeSegment, isSafeRelativePath, destination containment, same-origin URL checks). Add regression tests that assert hostile indexes do not fetch or write outside the skills cache.

Not a duplicate of #35996 (that PR covers local symlink walks under .claude/.agents; this covers remote index field validation).

How did you verify your code works?

• bun test --preload ./test/preload.ts test/skill/discovery.test.ts (hostile index cases included)

Screenshots / recordings

N/A — non-UI

Checklist

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

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Potential duplicate found:

• PR fix(skill): avoid symlink traversal during discovery #35996: fix(skill): avoid symlink traversal during discovery - fix(skill): avoid symlink traversal during discovery #35996

This PR appears related as it also addresses path traversal safety during skill discovery. You should verify if #35996 already covers the remote skill index scenario or if they address different aspects of skill discovery security.

[1837620622]

Review-ready

High-confidence path-traversal fix for remote skill discovery (aligns with packages/core guards). Tests included. Happy to iterate.

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍

[1837620622]

Friendly review ping 🙏

Path-traversal hardening for remote skill discovery (hostile skill.name / files in index.json). Aligns with the guards already used in packages/core. Tests included; checks green.

Happy to iterate if anything looks off.

— @1837620622