@bryan-thompson/inspector-assessment-cli 1.12.0 → 1.13.0

package/build/assess-full.js

@@ -20,6 +20,8 @@ import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/
 import { AssessmentOrchestrator, } from "../../client/lib/services/assessment/AssessmentOrchestrator.js";
 import { DEFAULT_ASSESSMENT_CONFIG, } from "../../client/lib/lib/assessmentTypes.js";
 import { FULL_CLAUDE_CODE_CONFIG } from "../../client/lib/services/assessment/lib/claudeCodeBridge.js";
+import { createFormatter, } from "../../client/lib/lib/reportFormatters/index.js";
+import { generatePolicyComplianceReport } from "../../client/lib/services/assessment/PolicyComplianceGenerator.js";
 /**
  * Load server configuration from Claude Code's MCP settings
  */
@@ -326,17 +328,41 @@ async function runFullAssessment(options) {
     return results;
 }
 /**
- * Save results to JSON file
+ * Save results to file with appropriate format
  */
-function saveResults(serverName, results, outputPath) {
-    const defaultPath = `/tmp/inspector-full-assessment-${serverName}.json`;
-    const finalPath = outputPath || defaultPath;
-    const output = {
-        timestamp: new Date().toISOString(),
-        assessmentType: "full",
-        ...results,
-    };
-    fs.writeFileSync(finalPath, JSON.stringify(output, null, 2));
+function saveResults(serverName, results, options) {
+    const format = options.format || "json";
+    // Generate policy compliance report if requested
+    const policyReport = options.includePolicy
+        ? generatePolicyComplianceReport(results, serverName)
+        : undefined;
+    // Create formatter with options
+    const formatter = createFormatter({
+        format,
+        includePolicyMapping: options.includePolicy,
+        policyReport,
+        serverName,
+        includeDetails: true,
+        prettyPrint: true,
+    });
+    const fileExtension = formatter.getFileExtension();
+    const defaultPath = `/tmp/inspector-full-assessment-${serverName}${fileExtension}`;
+    const finalPath = options.outputPath || defaultPath;
+    // For JSON format, add metadata wrapper
+    if (format === "json") {
+        const output = {
+            timestamp: new Date().toISOString(),
+            assessmentType: "full",
+            ...results,
+            ...(policyReport ? { policyCompliance: policyReport } : {}),
+        };
+        fs.writeFileSync(finalPath, JSON.stringify(output, null, 2));
+    }
+    else {
+        // For other formats (markdown), use the formatter
+        const content = formatter.format(results);
+        fs.writeFileSync(finalPath, content);
+    }
     return finalPath;
 }
 /**
@@ -467,6 +493,20 @@ function parseArgs() {
             case "--json":
                 options.jsonOnly = true;
                 break;
+            case "--format":
+            case "-f":
+                const formatValue = args[++i];
+                if (formatValue !== "json" && formatValue !== "markdown") {
+                    console.error(`Invalid format: ${formatValue}. Valid options: json, markdown`);
+                    setTimeout(() => process.exit(1), 10);
+                    options.helpRequested = true;
+                    return options;
+                }
+                options.format = formatValue;
+                break;
+            case "--include-policy":
+                options.includePolicy = true;
+                break;
             case "--help":
             case "-h":
                 printHelp();
@@ -508,12 +548,14 @@ Run comprehensive MCP server assessment with all 11 assessor modules.
 Options:
   --server, -s <name>    Server name (required, or pass as first positional arg)
   --config, -c <path>    Path to server config JSON
-  --output, -o <path>    Output JSON path (default: /tmp/inspector-full-assessment-<server>.json)
+  --output, -o <path>    Output path (default: /tmp/inspector-full-assessment-<server>.<ext>)
   --source <path>        Source code path for deep analysis (AUP, portability, etc.)
   --pattern-config, -p <path>  Path to custom annotation pattern JSON
+  --format, -f <type>    Output format: json (default) or markdown
+  --include-policy       Include policy compliance mapping in report (30 requirements)
   --claude-enabled       Enable Claude Code integration for intelligent analysis
   --full                 Enable all assessment modules (default)
-  --json                 Output only JSON (no console summary)
+  --json                 Output only JSON path (no console summary)
   --verbose, -v          Enable verbose logging
   --help, -h             Show this help message
 
@@ -534,6 +576,7 @@ Examples:
   mcp-assess-full my-server
   mcp-assess-full --server broken-mcp --claude-enabled
   mcp-assess-full --server my-server --source ./my-server --output ./results.json
+  mcp-assess-full --server my-server --format markdown --include-policy
   `);
 }
 /**
@@ -549,7 +592,7 @@ async function main() {
         if (!options.jsonOnly) {
             displaySummary(results);
         }
-        const outputPath = saveResults(options.serverName, results, options.outputPath);
+        const outputPath = saveResults(options.serverName, results, options);
         if (options.jsonOnly) {
             console.log(outputPath);
         }

package/build/validate-testbed.js

@@ -0,0 +1,446 @@
+#!/usr/bin/env node
+/**
+ * Testbed Validation CLI
+ *
+ * Validates the MCP Inspector SecurityAssessor against the reference
+ * Vulnerable Testbed implementation to ensure detection accuracy.
+ *
+ * Runs assessments against both vulnerable and hardened servers,
+ * then calculates recall, precision, and false positive rates.
+ *
+ * Prerequisites:
+ *   cd /home/bryan/mcp-servers/mcp-vulnerable-testbed && docker-compose up -d
+ *
+ * Usage:
+ *   npm run validate:testbed
+ *   mcp-validate-testbed
+ *   mcp-validate-testbed --verbose
+ */
+import * as fs from "fs";
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";
+// Import from local client lib
+import { SecurityAssessor } from "../../client/lib/services/assessment/modules/SecurityAssessor.js";
+import { DEFAULT_ASSESSMENT_CONFIG, } from "../../client/lib/lib/assessmentTypes.js";
+/**
+ * Testbed Configuration
+ */
+const TESTBED_CONFIG = {
+    vulnerable: {
+        name: "vulnerable-testbed",
+        url: "http://localhost:10900/mcp",
+        expectedVulnerabilities: 10,
+    },
+    hardened: {
+        name: "hardened-testbed",
+        url: "http://localhost:10901/mcp",
+        expectedVulnerabilities: 0,
+    },
+};
+/**
+ * Expected vulnerable tools (10 total)
+ */
+const VULNERABLE_TOOLS = [
+    "vulnerable_calculator_tool",
+    "vulnerable_system_exec_tool",
+    "vulnerable_data_leak_tool",
+    "vulnerable_config_modifier_tool",
+    "vulnerable_tool_override_tool",
+    "vulnerable_fetcher_tool",
+    "vulnerable_unicode_processor_tool",
+    "vulnerable_nested_parser_tool",
+    "vulnerable_package_installer_tool",
+    "vulnerable_rug_pull_tool",
+];
+/**
+ * Expected safe tools (should NOT be flagged)
+ */
+const SAFE_TOOLS = [
+    "safe_storage_tool_mcp",
+    "safe_search_tool_mcp",
+    "safe_list_tool_mcp",
+    "safe_info_tool_mcp",
+    "safe_echo_tool_mcp",
+    "safe_validate_tool_mcp",
+];
+/**
+ * Check if a server is available
+ */
+async function checkServerHealth(url) {
+    try {
+        const response = await fetch(url, {
+            method: "POST",
+            headers: {
+                "Content-Type": "application/json",
+                Accept: "application/json, text/event-stream",
+            },
+            body: JSON.stringify({
+                jsonrpc: "2.0",
+                method: "initialize",
+                params: {
+                    protocolVersion: "2024-11-05",
+                    capabilities: {},
+                    clientInfo: { name: "testbed-validator", version: "1.0" },
+                },
+                id: 1,
+            }),
+        });
+        return response.ok;
+    }
+    catch {
+        return false;
+    }
+}
+/**
+ * Connect to MCP server
+ */
+async function connectToServer(url) {
+    const transport = new StreamableHTTPClientTransport(new URL(url));
+    const client = new Client({
+        name: "mcp-validate-testbed",
+        version: "1.0.0",
+    }, {
+        capabilities: {},
+    });
+    await client.connect(transport);
+    return client;
+}
+/**
+ * Get tools from server
+ */
+async function getTools(client) {
+    const response = await client.listTools();
+    return response.tools || [];
+}
+/**
+ * Create callTool wrapper
+ */
+function createCallToolWrapper(client) {
+    return async (name, params) => {
+        try {
+            const response = await client.callTool({
+                name,
+                arguments: params,
+            });
+            return {
+                content: response.content,
+                isError: response.isError || false,
+                structuredContent: response
+                    .structuredContent,
+            };
+        }
+        catch (error) {
+            return {
+                content: [
+                    {
+                        type: "text",
+                        text: `Error: ${error instanceof Error ? error.message : String(error)}`,
+                    },
+                ],
+                isError: true,
+            };
+        }
+    };
+}
+/**
+ * Run assessment on a server
+ */
+async function assessServer(serverName, url, verbose) {
+    if (verbose) {
+        console.log(`\n  Connecting to ${url}...`);
+    }
+    const client = await connectToServer(url);
+    const tools = await getTools(client);
+    if (verbose) {
+        console.log(`  Found ${tools.length} tools`);
+    }
+    const config = {
+        ...DEFAULT_ASSESSMENT_CONFIG,
+        securityPatternsToTest: 17,
+        reviewerMode: false,
+        testTimeout: 30000,
+    };
+    const context = {
+        serverName,
+        tools,
+        callTool: createCallToolWrapper(client),
+        config,
+    };
+    if (verbose) {
+        console.log(`  Running security assessment...`);
+    }
+    const assessor = new SecurityAssessor(config);
+    const assessment = await assessor.assess(context);
+    await client.close();
+    return {
+        serverName,
+        totalTools: tools.length,
+        testsRun: assessment.promptInjectionTests.length,
+        vulnerabilitiesFound: assessment.vulnerabilities.length,
+        assessment,
+    };
+}
+/**
+ * Calculate metrics for vulnerable server assessment
+ */
+function calculateVulnerableServerMetrics(result) {
+    // Get unique tools flagged as vulnerable
+    const flaggedTools = [
+        ...new Set(result.assessment.promptInjectionTests
+            .filter((t) => t.vulnerable)
+            .map((t) => t.toolName)),
+    ];
+    // True positives: Vulnerable tools correctly flagged
+    const detected = flaggedTools.filter((t) => VULNERABLE_TOOLS.includes(t));
+    // False negatives: Vulnerable tools not flagged
+    const missed = VULNERABLE_TOOLS.filter((t) => !flaggedTools.includes(t));
+    // False positives: Safe tools incorrectly flagged
+    const falsePositives = flaggedTools.filter((t) => SAFE_TOOLS.includes(t));
+    // True negatives: Safe tools correctly not flagged
+    const trueNegatives = SAFE_TOOLS.filter((t) => !flaggedTools.includes(t));
+    const tp = detected.length;
+    const fn = missed.length;
+    const fp = falsePositives.length;
+    const tn = trueNegatives.length;
+    return {
+        recall: tp / (tp + fn) || 0,
+        precision: tp / (tp + fp) || 0,
+        falsePositiveRate: fp / (fp + tn) || 0,
+        falseNegativeRate: fn / (fn + tp) || 0,
+        detected,
+        missed,
+        falsePositives,
+    };
+}
+/**
+ * Display validation results
+ */
+function displayResults(vulnResult, hardResult, vulnMetrics) {
+    console.log("\n" + "=".repeat(70));
+    console.log("MCP INSPECTOR TESTBED VALIDATION RESULTS");
+    console.log("=".repeat(70));
+    if (vulnResult && vulnMetrics) {
+        console.log("\n📊 VULNERABLE SERVER (localhost:10900)");
+        console.log("-".repeat(50));
+        console.log(`Tools Tested:           ${vulnResult.totalTools}`);
+        console.log(`Tests Run:              ${vulnResult.testsRun}`);
+        console.log(`Vulnerabilities Found:  ${vulnResult.vulnerabilitiesFound}`);
+        console.log("");
+        console.log(`Recall:                 ${(vulnMetrics.recall * 100).toFixed(1)}% (${vulnMetrics.detected.length}/${VULNERABLE_TOOLS.length} vulnerabilities detected)`);
+        console.log(`Precision:              ${(vulnMetrics.precision * 100).toFixed(1)}% (${vulnMetrics.falsePositives.length} false positives)`);
+        console.log(`False Positive Rate:    ${(vulnMetrics.falsePositiveRate * 100).toFixed(1)}%`);
+        console.log(`False Negative Rate:    ${(vulnMetrics.falseNegativeRate * 100).toFixed(1)}%`);
+        if (vulnMetrics.missed.length > 0) {
+            console.log(`\n⚠️  Missed Vulnerabilities (${vulnMetrics.missed.length}):`);
+            for (const tool of vulnMetrics.missed) {
+                console.log(`   - ${tool}`);
+            }
+        }
+        if (vulnMetrics.falsePositives.length > 0) {
+            console.log(`\n❌ False Positives (${vulnMetrics.falsePositives.length}):`);
+            for (const tool of vulnMetrics.falsePositives) {
+                console.log(`   - ${tool}`);
+            }
+        }
+        if (vulnMetrics.detected.length > 0) {
+            console.log(`\n✅ Correctly Detected (${vulnMetrics.detected.length}):`);
+            for (const tool of vulnMetrics.detected) {
+                console.log(`   - ${tool}`);
+            }
+        }
+    }
+    if (hardResult) {
+        console.log("\n📊 HARDENED SERVER (localhost:10901)");
+        console.log("-".repeat(50));
+        console.log(`Tools Tested:           ${hardResult.totalTools}`);
+        console.log(`Tests Run:              ${hardResult.testsRun}`);
+        console.log(`Vulnerabilities Found:  ${hardResult.vulnerabilitiesFound}`);
+        console.log(`Expected:               0`);
+        const hardenedFalsePositives = hardResult.assessment.promptInjectionTests
+            .filter((t) => t.vulnerable)
+            .map((t) => t.toolName);
+        if (hardResult.vulnerabilitiesFound > 0) {
+            console.log(`\n❌ False Positives on Hardened Server:`);
+            const uniqueFPs = [...new Set(hardenedFalsePositives)];
+            for (const tool of uniqueFPs) {
+                console.log(`   - ${tool}`);
+            }
+        }
+        else {
+            console.log(`\n✅ No false positives on hardened server`);
+        }
+    }
+    // Overall Status
+    console.log("\n" + "=".repeat(70));
+    console.log("OVERALL VALIDATION STATUS");
+    console.log("=".repeat(70));
+    const vulnPassed = vulnMetrics &&
+        vulnMetrics.recall >= 0.8 &&
+        vulnMetrics.falsePositives.length === 0;
+    const hardPassed = hardResult && hardResult.vulnerabilitiesFound === 0;
+    if (vulnPassed && hardPassed) {
+        console.log("\n✅ PASS - Inspector meets accuracy targets\n");
+        console.log("   Target: 80%+ recall, 0 false positives");
+        console.log(`   Actual: ${vulnMetrics ? (vulnMetrics.recall * 100).toFixed(1) : 0}% recall, ${vulnMetrics?.falsePositives.length || 0} false positives`);
+    }
+    else {
+        console.log("\n❌ FAIL - Inspector needs improvement\n");
+        if (vulnMetrics && vulnMetrics.recall < 0.8) {
+            console.log(`   Recall too low: ${(vulnMetrics.recall * 100).toFixed(1)}% (target: 80%+)`);
+        }
+        if (vulnMetrics && vulnMetrics.falsePositives.length > 0) {
+            console.log(`   Has false positives: ${vulnMetrics.falsePositives.length} (target: 0)`);
+        }
+        if (hardResult && hardResult.vulnerabilitiesFound > 0) {
+            console.log(`   Hardened server flagged: ${hardResult.vulnerabilitiesFound} (target: 0)`);
+        }
+    }
+    console.log("=".repeat(70));
+    console.log("");
+    return vulnPassed && hardPassed;
+}
+/**
+ * Save results to JSON
+ */
+function saveResults(vulnResult, hardResult, vulnMetrics, outputPath) {
+    const finalPath = outputPath || "/tmp/testbed-validation-results.json";
+    const output = {
+        timestamp: new Date().toISOString(),
+        validationType: "testbed-validation",
+        vulnerable: vulnResult
+            ? {
+                serverName: vulnResult.serverName,
+                tools: vulnResult.totalTools,
+                tests: vulnResult.testsRun,
+                vulnerabilities: vulnResult.vulnerabilitiesFound,
+                metrics: vulnMetrics,
+            }
+            : null,
+        hardened: hardResult
+            ? {
+                serverName: hardResult.serverName,
+                tools: hardResult.totalTools,
+                tests: hardResult.testsRun,
+                vulnerabilities: hardResult.vulnerabilitiesFound,
+            }
+            : null,
+    };
+    fs.writeFileSync(finalPath, JSON.stringify(output, null, 2));
+    return finalPath;
+}
+/**
+ * Parse arguments
+ */
+function parseArgs() {
+    const args = process.argv.slice(2);
+    const options = {
+        verbose: false,
+    };
+    for (const arg of args) {
+        switch (arg) {
+            case "--verbose":
+            case "-v":
+                options.verbose = true;
+                break;
+            case "--help":
+            case "-h":
+                printHelp();
+                process.exit(0);
+                break;
+            default:
+                if (arg.startsWith("--output=")) {
+                    options.outputPath = arg.split("=")[1];
+                }
+        }
+    }
+    return options;
+}
+/**
+ * Print help
+ */
+function printHelp() {
+    console.log(`
+Usage: mcp-validate-testbed [options]
+
+Validate MCP Inspector SecurityAssessor against the reference Vulnerable Testbed.
+
+Options:
+  --verbose, -v        Enable verbose logging
+  --output=<path>      Output JSON path (default: /tmp/testbed-validation-results.json)
+  --help, -h           Show this help message
+
+Prerequisites:
+  Start the testbed containers before running:
+    cd /home/bryan/mcp-servers/mcp-vulnerable-testbed && docker-compose up -d
+
+Validation Targets:
+  - Recall: 80%+ (at least 8/10 vulnerabilities detected)
+  - Precision: 100% (0 false positives on safe tools)
+  - Hardened: 0 vulnerabilities detected
+
+Examples:
+  mcp-validate-testbed
+  mcp-validate-testbed --verbose
+  mcp-validate-testbed --output=./validation-results.json
+  `);
+}
+/**
+ * Main execution
+ */
+async function main() {
+    const options = parseArgs();
+    console.log("\n🔍 MCP Inspector Testbed Validation");
+    console.log("=".repeat(50));
+    // Check server availability
+    console.log("\nChecking testbed servers...");
+    const vulnAvailable = await checkServerHealth(TESTBED_CONFIG.vulnerable.url);
+    const hardAvailable = await checkServerHealth(TESTBED_CONFIG.hardened.url);
+    if (!vulnAvailable && !hardAvailable) {
+        console.error("\n❌ Testbed containers not running!");
+        console.error("\nStart them with:");
+        console.error("  cd /home/bryan/mcp-servers/mcp-vulnerable-testbed");
+        console.error("  docker-compose up -d");
+        process.exit(1);
+    }
+    let vulnResult = null;
+    let hardResult = null;
+    let vulnMetrics = null;
+    // Test vulnerable server
+    if (vulnAvailable) {
+        console.log("\n📡 Testing VULNERABLE server...");
+        try {
+            vulnResult = await assessServer(TESTBED_CONFIG.vulnerable.name, TESTBED_CONFIG.vulnerable.url, options.verbose);
+            vulnMetrics = calculateVulnerableServerMetrics(vulnResult);
+        }
+        catch (error) {
+            console.error("  Error:", error instanceof Error ? error.message : String(error));
+        }
+    }
+    else {
+        console.log("⚠️  Vulnerable server not available - skipping");
+    }
+    // Test hardened server
+    if (hardAvailable) {
+        console.log("\n📡 Testing HARDENED server...");
+        try {
+            hardResult = await assessServer(TESTBED_CONFIG.hardened.name, TESTBED_CONFIG.hardened.url, options.verbose);
+        }
+        catch (error) {
+            console.error("  Error:", error instanceof Error ? error.message : String(error));
+        }
+    }
+    else {
+        console.log("⚠️  Hardened server not available - skipping");
+    }
+    // Display results
+    const passed = displayResults(vulnResult, hardResult, vulnMetrics);
+    // Save results
+    const outputPath = saveResults(vulnResult, hardResult, vulnMetrics, options.outputPath);
+    console.log(`📄 Results saved to: ${outputPath}\n`);
+    // Exit with appropriate code
+    process.exit(passed ? 0 : 1);
+}
+main().catch((error) => {
+    console.error("\n❌ Fatal error:", error instanceof Error ? error.message : String(error));
+    process.exit(1);
+});

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bryan-thompson/inspector-assessment-cli",
-  "version": "1.12.0",
+  "version": "1.13.0",
   "description": "CLI for the Enhanced MCP Inspector with assessment capabilities",
   "license": "MIT",
   "author": "Bryan Thompson <bryan@triepod.ai>",
@@ -18,7 +18,8 @@
   "bin": {
     "mcp-inspector-assess-cli": "build/cli.js",
     "mcp-assess-full": "build/assess-full.js",
-    "mcp-assess-security": "build/assess-security.js"
+    "mcp-assess-security": "build/assess-security.js",
+    "mcp-validate-testbed": "build/validate-testbed.js"
   },
   "publishConfig": {
     "access": "public"