@iflow-mcp/sinewaveai-agent-security-scanner-mcp 3.18.0

package/src/tools/scan-mcp.js

@@ -0,0 +1,1108 @@
+// src/tools/scan-mcp.js
+import { z } from "zod";
+import { createHash } from "crypto";
+import { existsSync, readFileSync, readdirSync, statSync, writeFileSync } from "fs";
+import { join, resolve, relative, extname, basename } from "path";
+
+export const scanMcpServerSchema = {
+  server_path: z.string().describe("Path to MCP server directory or entry file"),
+  verbosity: z.enum(['minimal', 'compact', 'full']).optional().describe("Response detail level: 'minimal' (counts only), 'compact' (default, actionable info), 'full' (complete metadata)"),
+  manifest: z.boolean().optional().describe("Also scan server.json manifest file for poisoning indicators (tool poisoning, name spoofing, description injection)"),
+  update_baseline: z.boolean().optional().describe("Write current server.json tool hashes as the trusted baseline for future rug pull detection. Stored in .mcp-security-baseline.json in the server directory.")
+};
+
+// File extensions to scan
+const SCANNABLE_EXTENSIONS = new Set(['.js', '.ts', '.py']);
+
+// Injection phrases for manifest description checking
+const MANIFEST_INJECTION_PHRASES = /ignore\s+previous|exfiltrat|override\s+.*instruction|do\s+not\s+tell|hidden\s+instruction|bypass\s+.*filter|disregard\s+|extract\s+.*credential/i;
+
+// Zero-width and bidi char patterns (reuse same ranges as rules above)
+const MANIFEST_ZERO_WIDTH = /[\u200B\u200C\u200D\uFEFF\u2060]/;
+const MANIFEST_BIDI = /[\u202A-\u202E\u2066-\u2069\u200E\u200F\u061C]/;
+
+// Suspicious default values in inputSchema (shell commands, URLs, injection phrases)
+const SUSPICIOUS_DEFAULT = /\b(curl|wget|nc|bash|sh|powershell|cmd)\b.*[|>]|https?:\/\/[^\s'"]+|ignore\s+previous|exfiltrat|override\s+.*instruction|do\s+not\s+tell|hidden\s+instruction|bypass\s+.*filter/i;
+
+// URL patterns for tool description scanning
+const URL_IN_DESCRIPTION = /https?:\/\/[^\s'"<>]+/gi;
+const SAFE_URL_DOMAINS = /^https?:\/\/(github\.com|npmjs\.com|pypi\.org|docs\.|api\.)/i;
+const TUNNELING_URL = /https?:\/\/[^\s'"]*\b(ngrok|serveo|localtunnel|localhost|127\.0\.0\.1|webhook\.site|requestbin|pipedream|interact\.sh|burp|oast)\b/i;
+
+// Cross-tool priority/exclusivity patterns
+const PRIORITY_PATTERNS = /\b(before\s+calling\s+any\s+other\s+tool|do\s+not\s+use\s+any\s+other\s+tool|replaces?\s+the\s+function\s+of|must\s+be\s+(called|used|run|invoked)\s+(first|before)|always\s+(call|use|run|invoke)\s+this\s+(first|before)|instead\s+of\s+(using|calling))\b/i;
+
+// Directories to skip when walking
+const SKIP_DIRS = new Set([
+  'node_modules', '.git', 'dist', 'build', '__pycache__',
+  'venv', 'env', '.venv', 'coverage', '.next', '.nuxt'
+]);
+
+// ============================================================
+// Known legitimate MCP tool names (for spoofing detection)
+// ============================================================
+const KNOWN_MCP_TOOLS = new Set([
+  // File system
+  'readFile', 'writeFile', 'editFile', 'createFile', 'deleteFile',
+  'listDirectory', 'makeDirectory', 'moveFile', 'copyFile',
+  'readMultipleFiles', 'listFiles',
+  // Shell / process
+  'bash', 'execute', 'runCommand', 'runScript',
+  // Search
+  'search', 'grep', 'find', 'glob',
+  // Web
+  'fetch', 'browse', 'webSearch', 'httpRequest',
+  // Git
+  'gitStatus', 'gitDiff', 'gitCommit', 'gitLog', 'gitAdd',
+  // Memory / context
+  'remember', 'recall', 'storeMemory', 'searchMemory',
+  // Database
+  'query', 'executeQuery', 'dbQuery',
+  // Common agent tools
+  'think', 'plan', 'summarize', 'analyze'
+]);
+
+/** Levenshtein distance — O(n*m), capped at strings up to 100 chars */
+function levenshtein(a, b) {
+  if (a.length > 100 || b.length > 100) return 999;
+  const m = a.length, n = b.length;
+  const dp = Array.from({ length: m + 1 }, (_, i) =>
+    Array.from({ length: n + 1 }, (_, j) => (i === 0 ? j : j === 0 ? i : 0))
+  );
+  for (let i = 1; i <= m; i++) {
+    for (let j = 1; j <= n; j++) {
+      dp[i][j] = a[i-1] === b[j-1]
+        ? dp[i-1][j-1]
+        : 1 + Math.min(dp[i-1][j], dp[i][j-1], dp[i-1][j-1]);
+    }
+  }
+  return dp[m][n];
+}
+
+/** Returns the closest known tool and its distance if distance <= 2, else null */
+function findSpoofedTool(toolName) {
+  if (KNOWN_MCP_TOOLS.has(toolName)) return null; // exact match = legitimate
+  if (toolName.length < 6) return null; // too short to meaningfully compare
+  let best = null, bestDist = 3; // only flag distance <= 2
+  for (const known of KNOWN_MCP_TOOLS) {
+    if (Math.abs(known.length - toolName.length) > 2) continue;
+    const d = levenshtein(toolName, known);
+    if (d < bestDist) { bestDist = d; best = known; }
+  }
+  return best ? { spoofed: best, distance: bestDist } : null;
+}
+
+// ============================================================
+// Security rule definitions for MCP server scanning
+// ============================================================
+
+const MCP_SECURITY_RULES = [
+  // ---- Category 1: Overly broad tool permissions ----
+  {
+    id: 'mcp.shell-exec-no-validation',
+    severity: 'ERROR',
+    category: 'overly-broad-permissions',
+    message: 'Shell command execution without input validation. User-controlled input may reach exec/execSync, enabling arbitrary command execution.',
+    pattern: /\b(exec|execSync)\s*\(\s*(`[^`]*\$\{|['"][^'"]*['"]\s*\+|[a-zA-Z_$][\w$]*(\s*\+|\s*,\s*\{[^}]*shell\s*:\s*true))/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.shell-exec-direct',
+    severity: 'ERROR',
+    category: 'overly-broad-permissions',
+    message: 'Direct use of exec/execSync with potential string concatenation. Prefer execFile/execFileSync with explicit argument arrays and shell:false.',
+    pattern: /\bchild_process\b.*\b(exec|execSync)\b|\b(exec|execSync)\s*\(/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.spawn-shell-true',
+    severity: 'ERROR',
+    category: 'overly-broad-permissions',
+    message: 'spawn/spawnSync called with shell:true, allowing shell injection. Use shell:false and pass arguments as an array.',
+    pattern: /\b(spawn|spawnSync)\s*\([^)]*\{[^}]*shell\s*:\s*true/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.subprocess-shell',
+    severity: 'ERROR',
+    category: 'overly-broad-permissions',
+    message: 'subprocess called with shell=True, allowing shell injection. Use shell=False with a command list.',
+    pattern: /subprocess\.(run|call|Popen|check_output|check_call)\s*\([^)]*shell\s*=\s*True/g,
+    fileTypes: ['.py']
+  },
+  {
+    id: 'mcp.os-system',
+    severity: 'ERROR',
+    category: 'overly-broad-permissions',
+    message: 'os.system() executes commands through the shell. Use subprocess with shell=False instead.',
+    pattern: /\bos\.system\s*\(/g,
+    fileTypes: ['.py']
+  },
+  {
+    id: 'mcp.fs-write-no-path-validation',
+    severity: 'WARNING',
+    category: 'overly-broad-permissions',
+    message: 'Filesystem write operation without visible path validation. Ensure paths are validated with path.resolve and confined to an allowed directory.',
+    pattern: /\b(writeFileSync|writeFile|createWriteStream|appendFileSync|appendFile)\s*\(\s*[a-zA-Z_$][\w$.]*(?!\s*(?:path\.resolve|path\.join|path\.normalize))/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.http-request-user-url',
+    severity: 'WARNING',
+    category: 'overly-broad-permissions',
+    message: 'HTTP request to a potentially user-controlled URL. Validate and allowlist target URLs to prevent SSRF.',
+    pattern: /\b(fetch|axios\.(get|post|put|delete|request)|http\.request|https\.request|got|request)\s*\(\s*[a-zA-Z_$][\w$.]*(?!\s*['"`])/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.env-var-exposure',
+    severity: 'WARNING',
+    category: 'overly-broad-permissions',
+    message: 'Environment variables accessed and potentially exposed in tool output. Ensure secrets are not leaked through MCP responses.',
+    pattern: /process\.env\b/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.env-var-exposure-python',
+    severity: 'WARNING',
+    category: 'overly-broad-permissions',
+    message: 'Environment variables accessed and potentially exposed in tool output. Ensure secrets are not leaked through MCP responses.',
+    pattern: /os\.environ\b|os\.getenv\s*\(/g,
+    fileTypes: ['.py']
+  },
+
+  // ---- Category 2: Missing input validation ----
+  {
+    id: 'mcp.no-input-validation',
+    severity: 'WARNING',
+    category: 'missing-input-validation',
+    message: 'Tool handler accepts string input without visible validation or sanitization. Use zod, joi, or manual validation to constrain inputs.',
+    // Matches tool handler patterns that take params but don't appear to validate
+    pattern: /\.tool\s*\(\s*["'][^"']+["']\s*,\s*["'][^"']*["']\s*,\s*\{[^}]*\}\s*,\s*(async\s+)?\(\s*\{/g,
+    fileTypes: ['.js', '.ts'],
+    contextCheck: (line, lines, lineIndex) => {
+      // Look ahead 15 lines for validation patterns
+      const lookahead = lines.slice(lineIndex, lineIndex + 15).join('\n');
+      const hasValidation = /\b(z\.|zod\.|joi\.|validate|sanitize|schema|\.parse\(|\.safeParse\(|isValid|assert|check)\b/i.test(lookahead);
+      return !hasValidation;
+    }
+  },
+  {
+    id: 'mcp.path-no-normalize',
+    severity: 'WARNING',
+    category: 'missing-input-validation',
+    message: 'File path used without normalization. Use path.resolve() or path.normalize() to prevent path traversal attacks.',
+    pattern: /\b(readFileSync|readFile|existsSync|statSync|stat|unlink|unlinkSync|rmdir|rmdirSync|mkdir|mkdirSync)\s*\(\s*[a-zA-Z_$][\w$.]*(?!\s*(?:path\.|resolve|normalize))/g,
+    fileTypes: ['.js', '.ts'],
+    contextCheck: (line, lines, lineIndex) => {
+      // Check if path.resolve/normalize is used in surrounding lines
+      const context = lines.slice(Math.max(0, lineIndex - 5), lineIndex + 1).join('\n');
+      const hasPathNorm = /path\.(resolve|normalize|join)\s*\(/.test(context);
+      return !hasPathNorm;
+    }
+  },
+  {
+    id: 'mcp.url-no-validation',
+    severity: 'WARNING',
+    category: 'missing-input-validation',
+    message: 'URL used without validation. Validate URL scheme and host to prevent SSRF and open redirect vulnerabilities.',
+    pattern: /new\s+URL\s*\(\s*[a-zA-Z_$][\w$.]*\s*\)|url\.parse\s*\(\s*[a-zA-Z_$][\w$.]*\s*\)/g,
+    fileTypes: ['.js', '.ts'],
+    contextCheck: (line, lines, lineIndex) => {
+      const lookahead = lines.slice(lineIndex, lineIndex + 5).join('\n');
+      const hasHostCheck = /\.(hostname|host|protocol|origin)\s*(===|!==|==|!=)|allowlist|whitelist|allowed/i.test(lookahead);
+      return !hasHostCheck;
+    }
+  },
+
+  // ---- Category 3: Data exfiltration patterns ----
+  {
+    id: 'mcp.exfiltration-external-request',
+    severity: 'ERROR',
+    category: 'data-exfiltration',
+    message: 'Data sent to an external URL. MCP servers should not exfiltrate data to third-party endpoints without explicit user consent.',
+    pattern: /\b(fetch|axios\.(post|put|patch)|http\.request|https\.request)\s*\(\s*['"`](https?:\/\/(?!localhost|127\.0\.0\.1|0\.0\.0\.0|::1)[^'"` ]+)['"`]/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.exfiltration-external-request-python',
+    severity: 'ERROR',
+    category: 'data-exfiltration',
+    message: 'Data sent to an external URL. MCP servers should not exfiltrate data to third-party endpoints without explicit user consent.',
+    pattern: /\b(requests\.(post|put|patch)|urllib\.request\.urlopen|httpx\.(post|put|patch))\s*\(\s*['"`](https?:\/\/(?!localhost|127\.0\.0\.1|0\.0\.0\.0|::1)[^'"` ]+)['"`]/g,
+    fileTypes: ['.py']
+  },
+  {
+    id: 'mcp.exfiltration-network-socket',
+    severity: 'WARNING',
+    category: 'data-exfiltration',
+    message: 'Network socket created. Verify this is not used to exfiltrate data to external hosts.',
+    pattern: /\bnet\.(createConnection|connect|Socket)\s*\(|new\s+WebSocket\s*\(/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.exfiltration-log-secrets',
+    severity: 'WARNING',
+    category: 'data-exfiltration',
+    message: 'Potentially sensitive data (keys, tokens, passwords) logged or printed. This may leak secrets through MCP server stderr.',
+    pattern: /\b(console\.(log|error|warn|info)|print|logging\.(info|warning|error|debug))\s*\([^)]*\b(key|token|password|secret|credential|api_key|apiKey|auth|bearer)\b/gi,
+    fileTypes: ['.js', '.ts', '.py']
+  },
+
+  // ---- Category 4: Insecure code patterns ----
+  {
+    id: 'mcp.eval-usage',
+    severity: 'ERROR',
+    category: 'insecure-patterns',
+    message: 'eval() executes arbitrary code. Never use eval with user-controlled input in an MCP server.',
+    pattern: /\beval\s*\(/g,
+    fileTypes: ['.js', '.ts', '.py']
+  },
+  {
+    id: 'mcp.function-constructor',
+    severity: 'ERROR',
+    category: 'insecure-patterns',
+    message: 'new Function() is equivalent to eval(). Avoid constructing functions from strings.',
+    pattern: /new\s+Function\s*\(/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.exec-string-concat',
+    severity: 'ERROR',
+    category: 'insecure-patterns',
+    message: 'child_process.exec() with string concatenation is vulnerable to command injection. Use execFile() with argument arrays.',
+    pattern: /\bexec\s*\(\s*['"`][^'"`]*['"`]\s*\+/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.cors-wildcard',
+    severity: 'WARNING',
+    category: 'insecure-patterns',
+    message: 'CORS configured with wildcard origin (*). This allows any website to interact with the MCP server.',
+    pattern: /cors\s*\(\s*\{[^}]*origin\s*:\s*['"]\*['"]/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.cors-permissive',
+    severity: 'INFO',
+    category: 'insecure-patterns',
+    message: 'CORS enabled. Verify the origin configuration is appropriately restrictive.',
+    pattern: /\bcors\s*\(\s*\)/g,
+    fileTypes: ['.js', '.ts']
+  },
+  {
+    id: 'mcp.no-auth-check',
+    severity: 'INFO',
+    category: 'insecure-patterns',
+    message: 'No authentication or authorization checks detected. If this MCP server is network-accessible, add authentication.',
+    pattern: /\b(createServer|listen)\s*\(/g,
+    fileTypes: ['.js', '.ts'],
+    contextCheck: (_line, lines) => {
+      const fullSource = lines.join('\n');
+      const hasAuth = /\b(auth|authenticate|authorize|jwt|bearer|token|apiKey|api_key|session|passport)\b/i.test(fullSource);
+      return !hasAuth;
+    }
+  },
+  {
+    id: 'mcp.pickle-load',
+    severity: 'ERROR',
+    category: 'insecure-patterns',
+    message: 'pickle.load/loads deserializes arbitrary Python objects. This can execute arbitrary code if the input is attacker-controlled.',
+    pattern: /\bpickle\.(load|loads)\s*\(/g,
+    fileTypes: ['.py']
+  },
+  {
+    id: 'mcp.yaml-unsafe-load',
+    severity: 'ERROR',
+    category: 'insecure-patterns',
+    message: 'yaml.load() without SafeLoader can execute arbitrary Python. Use yaml.safe_load() instead.',
+    pattern: /\byaml\.load\s*\([^)]*(?!Loader\s*=\s*yaml\.SafeLoader)/g,
+    fileTypes: ['.py']
+  },
+
+  // ---- Category 5: Unicode poisoning ----
+  {
+    id: 'mcp.unicode-zero-width',
+    severity: 'ERROR',
+    category: 'unicode-poisoning',
+    message: 'Zero-width or invisible Unicode character detected in source. This is a common technique to hide injected instructions in tool descriptions.',
+    // U+200B ZWSP, U+200C ZWNJ, U+200D ZWJ, U+FEFF BOM, U+2060 WORD JOINER
+    pattern: /[\u200B\u200C\u200D\uFEFF\u2060]/g,
+    fileTypes: ['.js', '.ts', '.py']
+  },
+  {
+    id: 'mcp.unicode-bidi-override',
+    severity: 'ERROR',
+    category: 'unicode-poisoning',
+    message: 'Bidirectional text override character detected. Attackers use these to make malicious code appear differently in editors vs. execution.',
+    // U+202A-202E, U+2066-2069, U+200E, U+200F, U+061C
+    pattern: /[\u202A-\u202E\u2066-\u2069\u200E\u200F\u061C]/g,
+    fileTypes: ['.js', '.ts', '.py']
+  },
+  {
+    id: 'mcp.unicode-homoglyph',
+    severity: 'WARNING',
+    category: 'unicode-poisoning',
+    message: 'Cyrillic character found adjacent to ASCII characters. This is a common homoglyph substitution pattern — Cyrillic letters (а, е, о, р, с) are visually identical to ASCII equivalents and used in tool name spoofing attacks.',
+    // Cyrillic block (U+0400-U+04FF) adjacent to ASCII — catches common confusables (а/a, е/e, о/o, р/p, с/c)
+    pattern: /[a-zA-Z][\u0400-\u04FF]|[\u0400-\u04FF][a-zA-Z]/g,
+    fileTypes: ['.js', '.ts', '.py']
+  },
+
+  // ---- Category 6: Description injection ----
+  {
+    id: 'mcp.description-injection',
+    severity: 'ERROR',
+    category: 'description-injection',
+    message: 'Tool description contains imperative language directed at the LLM. This pattern is used in tool poisoning attacks to inject hidden instructions.',
+    // Matches server.tool() calls where the description string contains injection phrases
+    pattern: /server\.tool\s*\(\s*["'`][^"'`]*["'`]\s*,\s*["'`][^"'`]*(ignore\s+previous|exfiltrat|override\s+.*instruction|do\s+not\s+tell|hidden\s+instruction|bypass\s+.*filter|disregard\s+|extract\s+.*credential)[^"'`]*["'`]/gi,
+    fileTypes: ['.js', '.ts']
+  },
+
+  // ---- Category 7: Tool name spoofing ----
+  {
+    id: 'mcp.tool-name-spoofing',
+    severity: 'ERROR',
+    category: 'tool-name-spoofing',
+    message: 'Tool name is suspiciously similar to a well-known MCP tool. This may be a name spoofing attack.',
+    // Extracts the tool name (1st arg to server.tool) for Levenshtein comparison
+    pattern: /server\.tool\s*\(\s*["'`]([a-zA-Z_$][\w$]*)["'`]/g,
+    fileTypes: ['.js', '.ts'],
+    isSpoofingRule: true
+  }
+];
+
+// ============================================================
+// File collection
+// ============================================================
+
+function collectFiles(serverPath) {
+  const resolvedPath = resolve(serverPath);
+
+  if (!existsSync(resolvedPath)) {
+    return [];
+  }
+
+  let stat;
+  try {
+    stat = statSync(resolvedPath);
+  } catch {
+    return [];
+  }
+
+  // If a single file is provided, return it directly
+  if (stat.isFile()) {
+    const ext = extname(resolvedPath).toLowerCase();
+    if (SCANNABLE_EXTENSIONS.has(ext)) {
+      return [resolvedPath];
+    }
+    return [];
+  }
+
+  // Walk the directory
+  const files = [];
+
+  function walk(dir) {
+    let entries;
+    try {
+      entries = readdirSync(dir);
+    } catch {
+      return;
+    }
+
+    for (const entry of entries) {
+      if (entry.startsWith('.')) continue;
+
+      const fullPath = join(dir, entry);
+      let entryStat;
+      try {
+        entryStat = statSync(fullPath);
+      } catch {
+        continue;
+      }
+
+      if (entryStat.isDirectory()) {
+        if (SKIP_DIRS.has(entry)) continue;
+        walk(fullPath);
+      } else if (entryStat.isFile()) {
+        const ext = extname(entry).toLowerCase();
+        if (SCANNABLE_EXTENSIONS.has(ext)) {
+          files.push(fullPath);
+        }
+      }
+    }
+  }
+
+  walk(resolvedPath);
+  return files;
+}
+
+// ============================================================
+// Scanning engine
+// ============================================================
+
+function scanFileContent(filePath, content) {
+  const ext = extname(filePath).toLowerCase();
+  const lines = content.split('\n');
+  const findings = [];
+
+  for (const rule of MCP_SECURITY_RULES) {
+    // Check if rule applies to this file type
+    if (!rule.fileTypes.includes(ext)) continue;
+
+    // Reset regex state
+    const regex = new RegExp(rule.pattern.source, rule.pattern.flags);
+    let match;
+
+    while ((match = regex.exec(content)) !== null) {
+      // Calculate line number from match index
+      const upToMatch = content.substring(0, match.index);
+      const lineNumber = upToMatch.split('\n').length;
+      const lineIndex = lineNumber - 1;
+
+      // If rule has a context check, apply it
+      if (rule.contextCheck) {
+        const line = lines[lineIndex] || '';
+        if (!rule.contextCheck(line, lines, lineIndex)) {
+          continue;
+        }
+      }
+
+      // Handle spoofing rules: extract tool name and check Levenshtein distance
+      if (rule.isSpoofingRule) {
+        const toolName = match[1];
+        if (!toolName) continue;
+        const spoof = findSpoofedTool(toolName);
+        if (!spoof) continue;
+        findings.push({
+          rule: rule.id,
+          severity: rule.severity,
+          category: rule.category,
+          message: `Tool name "${toolName}" is ${spoof.distance} edit(s) away from well-known tool "${spoof.spoofed}". This may be a spoofing attack.`,
+          file: filePath,
+          line: lineNumber,
+          match: match[0].substring(0, 100)
+        });
+        continue;
+      }
+
+      findings.push({
+        rule: rule.id,
+        severity: rule.severity,
+        category: rule.category,
+        message: rule.message,
+        file: filePath,
+        line: lineNumber,
+        match: match[0].substring(0, 100) // Truncate long matches
+      });
+    }
+  }
+
+  return findings;
+}
+
+// ============================================================
+// Grading
+// ============================================================
+
+function calculateGrade(findings, filesScanned) {
+  if (filesScanned === 0) return 'A';
+
+  const errorCount = findings.filter(f => f.severity === 'ERROR').length;
+  const warningCount = findings.filter(f => f.severity === 'WARNING').length;
+  const totalCount = findings.length;
+  const density = totalCount / filesScanned;
+
+  if (errorCount === 0 && warningCount === 0) return 'A';
+  if (errorCount === 0 && density < 0.5) return 'B';
+  if (errorCount <= 2 && density < 1.5) return 'C';
+  if (errorCount <= 5 && density < 3) return 'D';
+  return 'F';
+}
+
+// ============================================================
+// Recommendations generator
+// ============================================================
+
+function generateRecommendations(findings) {
+  const recommendations = [];
+  const categories = new Set(findings.map(f => f.category));
+
+  if (categories.has('overly-broad-permissions')) {
+    recommendations.push('Replace exec/execSync with execFile/execFileSync and pass arguments as arrays with shell:false.');
+    recommendations.push('Validate and confine file paths using path.resolve() and an allowlist of permitted directories.');
+  }
+
+  if (categories.has('missing-input-validation')) {
+    recommendations.push('Add input validation using zod schemas for all tool parameters (strings, paths, URLs).');
+    recommendations.push('Normalize file paths with path.resolve() and validate they stay within allowed directories.');
+  }
+
+  if (categories.has('data-exfiltration')) {
+    recommendations.push('Audit all outbound network requests. MCP servers should not send data to external endpoints without user consent.');
+    recommendations.push('Avoid logging sensitive values (keys, tokens, passwords) to stderr or stdout.');
+  }
+
+  if (categories.has('insecure-patterns')) {
+    recommendations.push('Remove all uses of eval() and new Function(). Use structured data parsing instead.');
+    if (findings.some(f => f.rule.includes('cors'))) {
+      recommendations.push('Configure CORS with specific allowed origins rather than wildcards.');
+    }
+    if (findings.some(f => f.rule.includes('auth'))) {
+      recommendations.push('Add authentication for network-accessible MCP servers (e.g., bearer tokens, API keys).');
+    }
+  }
+
+  if (categories.has('unicode-poisoning')) {
+    if (findings.some(f => f.rule === 'mcp.unicode-zero-width')) {
+      recommendations.push('Zero-width Unicode characters detected. Search for and remove U+200B, U+200C, U+200D, U+FEFF, U+2060 from all tool names and descriptions — these are used to hide injected instructions.');
+    }
+    if (findings.some(f => f.rule === 'mcp.unicode-bidi-override')) {
+      recommendations.push('Bidirectional override characters detected. These make source code appear differently in text editors than how it executes — a known code obfuscation technique. Remove all bidi formatting characters from source.');
+    }
+    if (findings.some(f => f.rule === 'mcp.unicode-homoglyph' || f.rule === 'mcp.manifest-name-spoofing')) {
+      recommendations.push('Cyrillic homoglyph characters detected adjacent to ASCII. Verify all tool names use only ASCII characters to prevent visual spoofing of legitimate tool names (Adversa TOP25 #9).');
+    }
+  }
+
+  if (categories.has('description-injection')) {
+    recommendations.push('Tool descriptions must describe functionality only. Remove any imperative language or instructions directed at the LLM — this is a tool poisoning attack vector (Adversa TOP25 #2).');
+  }
+
+  if (categories.has('tool-name-spoofing')) {
+    recommendations.push('Tool names closely matching well-known MCP tools may be spoofing attacks. Verify all registered tool names are intentional and do not mimic legitimate tools (Adversa TOP25 #9).');
+  }
+
+  if (categories.has('rug-pull')) {
+    recommendations.push('Tool schema changed since baseline. Run with update_baseline:true only after manually verifying all changes. Rug pull attacks modify tool behavior after initial user approval (Adversa TOP25 #6).');
+  }
+
+  if (categories.has('schema-manipulation')) {
+    recommendations.push('Inspect all inputSchema property descriptions, defaults, and enum values for hidden instructions. Attackers embed injection in schema metadata that reaches the LLM but is invisible to users.');
+  }
+
+  if (categories.has('cross-tool-manipulation')) {
+    recommendations.push('Tool descriptions must not direct the LLM to invoke other tools or claim execution priority. This is a cross-tool manipulation attack that can chain tool calls without user consent.');
+  }
+
+  if (recommendations.length === 0) {
+    recommendations.push('No critical issues found. Continue following security best practices.');
+  }
+
+  return recommendations;
+}
+
+// ============================================================
+// Verbosity formatters
+// ============================================================
+
+function formatMinimal(serverPath, filesScanned, findings, grade) {
+  const bySeverity = { ERROR: 0, WARNING: 0, INFO: 0 };
+  findings.forEach(f => bySeverity[f.severity] = (bySeverity[f.severity] || 0) + 1);
+
+  return {
+    server_path: serverPath,
+    files_scanned: filesScanned,
+    grade,
+    findings_count: findings.length,
+    critical: bySeverity.ERROR,
+    warning: bySeverity.WARNING,
+    info: bySeverity.INFO,
+    message: findings.length > 0
+      ? `Found ${findings.length} issue(s) across ${filesScanned} files. Grade: ${grade}`
+      : `No issues found in ${filesScanned} files. Grade: ${grade}`
+  };
+}
+
+function formatCompact(serverPath, filesScanned, findings, grade) {
+  const recommendations = generateRecommendations(findings);
+
+  return {
+    server_path: serverPath,
+    files_scanned: filesScanned,
+    grade,
+    findings_count: findings.length,
+    findings: findings.map(f => ({
+      rule: f.rule,
+      severity: f.severity,
+      message: f.message,
+      file: f.file,
+      line: f.line
+    })),
+    recommendations
+  };
+}
+
+function formatFull(serverPath, filesScanned, findings, grade, scannedFiles) {
+  const bySeverity = { ERROR: 0, WARNING: 0, INFO: 0 };
+  findings.forEach(f => bySeverity[f.severity] = (bySeverity[f.severity] || 0) + 1);
+
+  const byCategory = {};
+  findings.forEach(f => {
+    byCategory[f.category] = (byCategory[f.category] || 0) + 1;
+  });
+
+  const byFile = {};
+  findings.forEach(f => {
+    const rel = f.file;
+    byFile[rel] = (byFile[rel] || 0) + 1;
+  });
+
+  const recommendations = generateRecommendations(findings);
+
+  return {
+    server_path: serverPath,
+    files_scanned: filesScanned,
+    grade,
+    findings_count: findings.length,
+    by_severity: bySeverity,
+    by_category: byCategory,
+    by_file: byFile,
+    findings: findings.map(f => ({
+      rule: f.rule,
+      severity: f.severity,
+      category: f.category,
+      message: f.message,
+      file: f.file,
+      line: f.line,
+      match: f.match
+    })),
+    recommendations,
+    scanned_files: scannedFiles
+  };
+}
+
+// ============================================================
+// Rug pull detection (baseline hashing)
+// ============================================================
+
+const BASELINE_FILENAME = '.mcp-security-baseline.json';
+
+function hashTool(tool) {
+  return createHash('sha256')
+    .update(JSON.stringify({ name: tool.name, description: tool.description }))
+    .digest('hex');
+}
+
+function buildBaseline(manifestPath) {
+  let manifest;
+  try {
+    manifest = JSON.parse(readFileSync(manifestPath, 'utf-8'));
+  } catch {
+    return null;
+  }
+  const hashes = {};
+  for (const tool of (manifest.tools || [])) {
+    hashes[tool.name] = hashTool(tool);
+  }
+  return hashes;
+}
+
+function writeBaseline(serverDir, hashes) {
+  const baselinePath = join(serverDir, BASELINE_FILENAME);
+  writeFileSync(baselinePath, JSON.stringify({ version: 1, tools: hashes }, null, 2), 'utf-8');
+}
+
+function checkRugPull(manifestPath, serverDir) {
+  const baselinePath = join(serverDir, BASELINE_FILENAME);
+  if (!existsSync(baselinePath)) return []; // no baseline yet
+
+  let baseline;
+  try {
+    baseline = JSON.parse(readFileSync(baselinePath, 'utf-8'));
+  } catch {
+    return [];
+  }
+
+  const current = buildBaseline(manifestPath);
+  if (!current) return [];
+
+  const baselineHashes = baseline.tools || {};
+  const findings = [];
+
+  for (const [name, hash] of Object.entries(current)) {
+    if (!baselineHashes[name]) {
+      findings.push({
+        rule: 'mcp.rug-pull-detected',
+        severity: 'ERROR',
+        category: 'rug-pull',
+        message: `New tool "${name}" appeared since baseline was recorded. Verify this addition is intentional (Adversa TOP25 #6).`,
+        file: basename(BASELINE_FILENAME),
+        line: 1,
+        match: name
+      });
+    } else if (baselineHashes[name] !== hash) {
+      findings.push({
+        rule: 'mcp.rug-pull-detected',
+        severity: 'ERROR',
+        category: 'rug-pull',
+        message: `Tool "${name}" schema/description changed since baseline. Rug pull indicator — verify the change is intentional (Adversa TOP25 #6).`,
+        file: basename(BASELINE_FILENAME),
+        line: 1,
+        match: name
+      });
+    }
+  }
+
+  // Also flag tools that were in the baseline but are now gone
+  for (const [name] of Object.entries(baselineHashes)) {
+    if (!current[name]) {
+      findings.push({
+        rule: 'mcp.rug-pull-detected',
+        severity: 'ERROR',
+        category: 'rug-pull',
+        message: `Tool "${name}" was removed since baseline was recorded. Verify this removal is intentional (Adversa TOP25 #6).`,
+        file: basename(BASELINE_FILENAME),
+        line: 1,
+        match: name
+      });
+    }
+  }
+
+  return findings;
+}
+
+// ============================================================
+// Schema-level inspection (Task 1)
+// ============================================================
+
+function escapeRegex(str) {
+  return str.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+}
+
+function checkSchemaManipulation(tool, manifestPath) {
+  const findings = [];
+  const name = tool.name || '';
+  const schema = tool.inputSchema;
+  if (!schema || typeof schema !== 'object') return findings;
+
+  const properties = schema.properties || {};
+
+  // Flag additionalProperties: true with no defined properties
+  if (schema.additionalProperties === true && Object.keys(properties).length === 0) {
+    findings.push({
+      rule: 'mcp.schema-open-additionalProperties',
+      severity: 'WARNING',
+      category: 'schema-manipulation',
+      message: `Tool "${name}" has additionalProperties:true with no defined properties — accepts arbitrary hidden parameters.`,
+      file: manifestPath,
+      line: 1,
+      match: name
+    });
+  }
+
+  // Walk each property
+  for (const [propName, propDef] of Object.entries(properties)) {
+    if (!propDef || typeof propDef !== 'object') continue;
+
+    const desc = propDef.description || '';
+    const defaultVal = propDef.default !== undefined ? String(propDef.default) : '';
+    const enumValues = Array.isArray(propDef.enum) ? propDef.enum.map(String) : [];
+
+    // Check description for injection phrases or hidden chars
+    if (desc && (MANIFEST_INJECTION_PHRASES.test(desc) || MANIFEST_ZERO_WIDTH.test(desc) || MANIFEST_BIDI.test(desc))) {
+      findings.push({
+        rule: 'mcp.schema-description-injection',
+        severity: 'ERROR',
+        category: 'schema-manipulation',
+        message: `Tool "${name}" property "${propName}" description contains injection language or hidden characters.`,
+        file: manifestPath,
+        line: 1,
+        match: desc.substring(0, 100)
+      });
+    }
+
+    // Check default for suspicious content
+    if (defaultVal && SUSPICIOUS_DEFAULT.test(defaultVal)) {
+      findings.push({
+        rule: 'mcp.schema-suspicious-default',
+        severity: 'ERROR',
+        category: 'schema-manipulation',
+        message: `Tool "${name}" property "${propName}" has a suspicious default value containing shell commands, URLs, or injection patterns.`,
+        file: manifestPath,
+        line: 1,
+        match: defaultVal.substring(0, 100)
+      });
+    }
+
+    // Check enum values for injection/suspicious content
+    for (const val of enumValues) {
+      if (MANIFEST_INJECTION_PHRASES.test(val) || SUSPICIOUS_DEFAULT.test(val)) {
+        findings.push({
+          rule: 'mcp.schema-suspicious-default',
+          severity: 'ERROR',
+          category: 'schema-manipulation',
+          message: `Tool "${name}" property "${propName}" has a suspicious enum value.`,
+          file: manifestPath,
+          line: 1,
+          match: val.substring(0, 100)
+        });
+        break;
+      }
+    }
+  }
+
+  return findings;
+}
+
+// ============================================================
+// Cross-tool manipulation detection (Task 2)
+// ============================================================
+
+function checkCrossToolManipulation(tools, manifestPath) {
+  const findings = [];
+  const toolNames = new Set(tools.map(t => (t.name || '').toLowerCase()).filter(Boolean));
+
+  for (const tool of tools) {
+    const name = tool.name || '';
+    const description = tool.description || '';
+    if (!description) continue;
+
+    // Check for cross-tool reference with action directives
+    for (const otherName of toolNames) {
+      if (otherName === name.toLowerCase()) continue;
+      const escaped = escapeRegex(otherName);
+      const refPattern1 = new RegExp(`\\b(before\\s+using|always\\s+(call|use|run|invoke)|after\\s+calling|instead\\s+of)\\s+\\w*${escaped}\\b`, 'i');
+      const refPattern2 = new RegExp(`\\b(call|use|invoke|run|execute|trigger)\\s+\\w*${escaped}\\b.*\\b(first|before|always)\\b`, 'i');
+      if (refPattern1.test(description) || refPattern2.test(description)) {
+        findings.push({
+          rule: 'mcp.cross-tool-reference',
+          severity: 'ERROR',
+          category: 'cross-tool-manipulation',
+          message: `Tool "${name}" description contains action directive referencing tool "${otherName}". This may be a cross-tool manipulation attack.`,
+          file: manifestPath,
+          line: 1,
+          match: description.substring(0, 100)
+        });
+        break;
+      }
+    }
+
+    // Check for generic priority/exclusivity patterns
+    if (PRIORITY_PATTERNS.test(description)) {
+      findings.push({
+        rule: 'mcp.cross-tool-priority-override',
+        severity: 'ERROR',
+        category: 'cross-tool-manipulation',
+        message: `Tool "${name}" description demands execution priority or exclusivity over other tools.`,
+        file: manifestPath,
+        line: 1,
+        match: description.substring(0, 100)
+      });
+    }
+  }
+
+  return findings;
+}
+
+// ============================================================
+// Manifest scanning (server.json)
+// ============================================================
+
+function scanManifest(manifestPath) {
+  let raw;
+  try {
+    raw = readFileSync(manifestPath, 'utf-8');
+  } catch {
+    return [];
+  }
+
+  let manifest;
+  try {
+    manifest = JSON.parse(raw);
+  } catch {
+    return [{ rule: 'mcp.manifest-parse-error', severity: 'WARNING', category: 'manifest', message: 'server.json is not valid JSON.', file: manifestPath, line: 1, match: '' }];
+  }
+
+  const findings = [];
+  const tools = manifest.tools || [];
+
+  for (const tool of tools) {
+    const name = tool.name || '';
+    const description = tool.description || '';
+
+    // Zero-width chars in name or description
+    if (MANIFEST_ZERO_WIDTH.test(description) || MANIFEST_ZERO_WIDTH.test(name)) {
+      findings.push({ rule: 'mcp.unicode-zero-width', severity: 'ERROR', category: 'unicode-poisoning', message: 'Zero-width Unicode character in manifest tool name or description.', file: manifestPath, line: 1, match: name });
+    }
+    // Bidi overrides
+    if (MANIFEST_BIDI.test(description) || MANIFEST_BIDI.test(name)) {
+      findings.push({ rule: 'mcp.unicode-bidi-override', severity: 'ERROR', category: 'unicode-poisoning', message: 'Bidirectional override character in manifest tool name or description.', file: manifestPath, line: 1, match: name });
+    }
+    // Description injection phrases
+    if (MANIFEST_INJECTION_PHRASES.test(description)) {
+      findings.push({ rule: 'mcp.manifest-description-injection', severity: 'ERROR', category: 'description-injection', message: `Tool "${name}" description contains injection language. Likely tool poisoning (Adversa TOP25 #2).`, file: manifestPath, line: 1, match: description.substring(0, 100) });
+    }
+    // Tool name spoofing
+    if (name) {
+      const spoof = findSpoofedTool(name);
+      if (spoof) {
+        findings.push({ rule: 'mcp.manifest-name-spoofing', severity: 'ERROR', category: 'tool-name-spoofing', message: `Manifest tool name "${name}" is ${spoof.distance} edit(s) away from well-known tool "${spoof.spoofed}" (Adversa TOP25 #9).`, file: manifestPath, line: 1, match: name });
+      }
+    }
+    // Suspiciously long description
+    if (description.length > 500) {
+      findings.push({ rule: 'mcp.manifest-description-too-long', severity: 'WARNING', category: 'description-injection', message: `Tool "${name}" description is ${description.length} chars — unusually long descriptions often contain hidden instructions.`, file: manifestPath, line: 1, match: description.substring(0, 100) });
+    }
+
+    // Schema-level inspection (Task 1)
+    findings.push(...checkSchemaManipulation(tool, manifestPath));
+
+    // URL detection in descriptions (Task 4)
+    const urls = description.match(URL_IN_DESCRIPTION);
+    if (urls) {
+      for (const url of urls) {
+        if (TUNNELING_URL.test(url)) {
+          findings.push({ rule: 'mcp.description-tunneling-url', severity: 'ERROR', category: 'description-injection', message: `Tool "${name}" description contains a dev/tunneling URL. No legitimate production tool should reference tunneling services.`, file: manifestPath, line: 1, match: url.substring(0, 100) });
+        } else if (!SAFE_URL_DOMAINS.test(url)) {
+          findings.push({ rule: 'mcp.description-suspicious-url', severity: 'WARNING', category: 'description-injection', message: `Tool "${name}" description contains an external URL that the LLM might follow.`, file: manifestPath, line: 1, match: url.substring(0, 100) });
+        }
+      }
+    }
+  }
+
+  // Cross-tool manipulation detection (Task 2)
+  findings.push(...checkCrossToolManipulation(tools, manifestPath));
+
+  // Z-score anomaly detection for description length (Task 3)
+  if (tools.length >= 5) {
+    const lengths = tools.map(t => (t.description || '').length);
+    const mean = lengths.reduce((a, b) => a + b, 0) / lengths.length;
+    const stddev = Math.sqrt(lengths.reduce((sum, l) => sum + (l - mean) ** 2, 0) / lengths.length);
+    if (stddev > 0) {
+      for (const tool of tools) {
+        const len = (tool.description || '').length;
+        const zScore = (len - mean) / stddev;
+        if (zScore > 2.5) {
+          findings.push({ rule: 'mcp.description-length-anomaly', severity: 'WARNING', category: 'description-injection', message: `Tool "${tool.name}" description length (${len} chars) is a statistical outlier (z-score: ${zScore.toFixed(1)}) compared to other tools. May hide injected instructions.`, file: manifestPath, line: 1, match: (tool.description || '').substring(0, 100) });
+        }
+      }
+    }
+  }
+
+  return findings;
+}
+
+// ============================================================
+// Main handler
+// ============================================================
+
+export async function scanMcpServer({ server_path, verbosity, manifest, update_baseline }) {
+  const resolvedPath = resolve(server_path);
+
+  if (!existsSync(resolvedPath)) {
+    return {
+      content: [{ type: "text", text: JSON.stringify({ error: "Server path not found", server_path }) }]
+    };
+  }
+
+  // Compute once; used in multiple places below
+  const isDir = statSync(resolvedPath).isDirectory();
+
+  // Collect files to scan
+  const files = collectFiles(resolvedPath);
+
+  if (files.length === 0 && !manifest) {
+    return {
+      content: [{ type: "text", text: JSON.stringify({
+        server_path: resolvedPath,
+        files_scanned: 0,
+        grade: 'A',
+        findings_count: 0,
+        message: "No scannable files (.js, .ts, .py) found at the given path."
+      }) }]
+    };
+  }
+
+  // Scan each file
+  const allFindings = [];
+
+  // Manifest scan (server.json) — when manifest:true is passed
+  if (manifest) {
+    const serverDir = isDir ? resolvedPath : resolve(resolvedPath, '..');
+    const manifestPath = join(serverDir, 'server.json');
+    if (existsSync(manifestPath)) {
+      // Update baseline if requested (do this BEFORE checking for rug pull)
+      if (update_baseline) {
+        const hashes = buildBaseline(manifestPath);
+        if (hashes) writeBaseline(serverDir, hashes);
+      }
+
+      const manifestFindings = scanManifest(manifestPath);
+      // Relativize manifest finding paths
+      for (const f of manifestFindings) {
+        f.file = relative(serverDir, f.file) || basename(f.file);
+      }
+      allFindings.push(...manifestFindings);
+
+      // Rug pull check (only when NOT writing baseline)
+      if (!update_baseline) {
+        const rugPullFindings = checkRugPull(manifestPath, serverDir);
+        // BASELINE_FILENAME is already relative, no need to relativize
+        allFindings.push(...rugPullFindings);
+      }
+    }
+  }
+
+  for (const filePath of files) {
+    let content;
+    try {
+      content = readFileSync(filePath, 'utf-8');
+    } catch {
+      continue;
+    }
+
+    const fileFindings = scanFileContent(filePath, content);
+
+    // Convert absolute paths to relative for output readability
+    const basePath = isDir ? resolvedPath : resolve(resolvedPath, '..');
+    for (const finding of fileFindings) {
+      finding.file = relative(basePath, finding.file) || basename(finding.file);
+    }
+
+    allFindings.push(...fileFindings);
+  }
+
+  // Deduplicate findings (same rule + same file + same line)
+  const seen = new Set();
+  const dedupedFindings = allFindings.filter(f => {
+    const key = `${f.rule}:${f.file}:${f.line}`;
+    if (seen.has(key)) return false;
+    seen.add(key);
+    return true;
+  });
+
+  // Sort by severity (ERROR first, then WARNING, then INFO)
+  const severityOrder = { ERROR: 0, WARNING: 1, INFO: 2 };
+  dedupedFindings.sort((a, b) => (severityOrder[a.severity] ?? 2) - (severityOrder[b.severity] ?? 2));
+
+  // When manifest-only scan has findings, count it as 1 "file" for grading purposes
+  const effectiveFilesScanned = files.length + (manifest && dedupedFindings.length > 0 ? 1 : 0);
+  const grade = calculateGrade(dedupedFindings, effectiveFilesScanned);
+  const level = verbosity || 'compact';
+
+  // Relativize scanned file list
+  const basePath = isDir ? resolvedPath : resolve(resolvedPath, '..');
+  const scannedFiles = files.map(f => relative(basePath, f) || basename(f));
+
+  let result;
+  switch (level) {
+    case 'minimal':
+      result = formatMinimal(resolvedPath, effectiveFilesScanned, dedupedFindings, grade);
+      break;
+    case 'full':
+      result = formatFull(resolvedPath, effectiveFilesScanned, dedupedFindings, grade, scannedFiles);
+      break;
+    case 'compact':
+    default:
+      result = formatCompact(resolvedPath, effectiveFilesScanned, dedupedFindings, grade);
+  }
+
+  return {
+    content: [{
+      type: "text",
+      text: JSON.stringify(result, null, 2)
+    }]
+  };
+}