Blind-Auditor-MCP

Instructions

Implementation Reference

• src/main.py:188-214 (handler)

  The complete handler function for the 'submit_audit_result' tool. It is registered via the @mcp.tool() decorator. The function signature provides the input schema (passed: bool, issues: list[str], score: int=0). It validates the score against a minimum threshold of 80, updates the session audit history, and returns success or failure messages based on the passed parameter.

  @mcp.tool()
  def submit_audit_result(passed: bool, issues: list[str], score: int = 0) -> str:
      """Submit the audit result."""
      print(f"DEBUG: submit_audit_result called: passed={passed}, score={score}", file=sys.stderr)
      
      # Hardcoded score validation
      MIN_SCORE = 80
      if passed and score < MIN_SCORE:
          passed = False
          issues.append(f"[SYSTEM ENFORCEMENT] Score ({score}) is below minimum threshold ({MIN_SCORE}). You cannot pass code with such a low score.")
      
      session.audit_history.append({
          "passed": passed,
          "issues": issues,
          "score": score,
          "retry_count": session.retry_count
      })
      
      if passed:
          session.status = "APPROVED"
          return f"✅ AUDIT PASSED (Score: {score}/100)\n\n```\n{session.current_code}\n```"
      else:
          session.retry_count += 1
          session.status = "IDLE"
          issues_formatted = "\n".join([f"- {issue}" for issue in issues])
          return f"❌ AUDIT FAILED (Score: {score}/100)\n\n**Issues:**\n{issues_formatted}\n\nRetry count: {session.retry_count}/{rules_loader.get_max_retries()}"