unity-compile-fixer

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEDATA_EXFILTRATION
Full Analysis
  • Indirect Prompt Injection (LOW): The skill processes untrusted diagnostic data from external tools (VSCode/OmniSharp) and interpolates this into its reasoning process for generating code fixes.
  • Ingestion points: mcp__ide__getDiagnostics tool output and file contents via the Read tool.
  • Boundary markers: The skill uses the AskUserQuestion tool to require human approval before applying any fixes, significantly mitigating the risk of automated malicious actions.
  • Capability inventory: The skill can modify local files using the Edit tool and execute a local Node.js analysis script.
  • Sanitization: No explicit sanitization or instruction-ignoring delimiters are defined for the incoming diagnostic messages.
  • Data Exposure (SAFE): The provided Node.js script reads a file path from command-line arguments. While it could be pointed at sensitive JSON files, it only outputs data that conforms to a specific VSCode diagnostic schema, and the JSON.parse step would fail for most sensitive system files (like SSH keys or passwd) which are not JSON-formatted.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 05:57 PM