chore(llm): Better defense against prompt injection in triage skill

[github-actions]

Note

The pull request "chore(llm): Better defense against prompt injection in triage skill" was created by @chargome but did not reference an issue. Therefore this issue was created for better visibility in external tools like Linear.

Adds

• Language filter: Reject non-English issues (detects accented characters)
• Injection detection: Scan for malicious patterns with confidence scoring

[linear]

JS-1757