Sanitize paths in produced json

[JanKrivanek]

Context

LLM Judge can be adding paths to result json that cause it to be unparseable - e.g.:

https://github.com/dotnet/skills/actions/runs/22298025045/job/64498823613#step:6:61

So sanitizing paths to prevent results being flagged as wrong

[Copilot]

Pull request overview

Adds a small JSON sanitization step in the skill-validator’s LLM judge parsing path to prevent invalid escape sequences (commonly from Windows-style paths) from breaking JSON.parse, and introduces unit tests to cover these cases.

Changes:

• Add sanitizeJsonEscapes() utility to repair invalid JSON escape sequences inside JSON strings.
• Apply sanitization before JSON.parse in both the independent judge and pairwise judge parsers.
• Add Vitest coverage for common valid/invalid escape scenarios and ignore generated validator results output.

Reviewed changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
eng/skill-validator/src/json-utils.ts	Introduces sanitizeJsonEscapes() to rewrite invalid escapes to valid JSON escapes.
eng/skill-validator/src/judge.ts	Sanitizes extracted JSON before parsing and reports sanitized JSON on parse errors.
eng/skill-validator/src/pairwise-judge.ts	Sanitizes extracted JSON before parsing in pairwise judge flow.
eng/skill-validator/tests/json-utils.test.ts	Adds unit tests covering valid escapes, invalid path escapes, and multiline judge output.
.gitignore	Ignores skill-validator’s default .skill-validator-results/ output directory.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[JanKrivanek]

Fixed by other PRs