@skyramp/mcp 0.0.46 → 0.0.48

package/build/index.js

@@ -127,19 +127,26 @@ async function main() {
     await server.connect(transport);
     logger.info("MCP Server started successfully");
     // Listen for stdin closure (parent process disconnected)
-    process.stdin.on("end", () => {
-        logger.info("STDIN closed, parent process disconnected. Exiting...");
-        process.exit(0);
+    // Using multiple events for robustness across different disconnection scenarios
+    let isShuttingDown = false;
+    const handleParentDisconnect = (event) => {
+        if (isShuttingDown)
+            return;
+        isShuttingDown = true;
+        logger.info(`Parent process disconnected (${event}). Cleaning up and exiting...`);
+        server.close().finally(() => {
+            process.exit(0);
+        });
+    };
+    process.stdin.on("end", () => handleParentDisconnect("stdin end"));
+    process.stdin.on("close", () => handleParentDisconnect("stdin close"));
+    process.stdin.on("error", (err) => {
+        logger.error("STDIN error, parent likely disconnected", { error: err });
+        handleParentDisconnect("stdin error");
     });
     // Handle process termination signals
-    process.on("SIGTERM", () => {
-        logger.info("Received SIGTERM, shutting down gracefully...");
-        process.exit(0);
-    });
-    process.on("SIGINT", () => {
-        logger.info("Received SIGINT, shutting down gracefully...");
-        process.exit(0);
-    });
+    process.on("SIGTERM", () => handleParentDisconnect("SIGTERM"));
+    process.on("SIGINT", () => handleParentDisconnect("SIGINT"));
 }
 main().catch(async (error) => {
     logger.critical("Fatal error in main()", {

package/build/prompts/test-recommendation/repository-analysis-prompt.js

@@ -282,6 +282,12 @@ VALIDATION CHECKLIST:
 - [ ] Infrastructure flags verified
 - [ ] Existing tests catalogued
 
-Begin analysis now. Return ONLY the JSON object, no other text.
+**CRITICAL INSTRUCTIONS**:
+- Return the JSON object directly in your response text.
+- DO NOT create or save any files (no repository_analysis.json or .md files).
+- DO NOT use file write tools or save analysis to disk.
+- Output the complete analysis JSON inline in your response.
+
+Begin analysis now. Return the JSON object directly in your response text.
 `;
 }

package/build/prompts/test-recommendation/test-recommendation-prompt.js

@@ -120,6 +120,12 @@ Return a JSON object with this structure:
 4. **Provide Guidance**: Clear instructions for creating missing artifacts
 5. **Include Evidence**: Reference specific repository characteristics in rationale
 
-Generate recommendations now. Return ONLY the JSON object, no other text.
+**CRITICAL INSTRUCTIONS**:
+- Return the JSON object directly in your response text.
+- DO NOT create or save any files (no test_recommendations.md, .json, or other files).
+- DO NOT use file write tools or save recommendations to disk.
+- Output the complete recommendations JSON inline in your response.
+
+Generate recommendations now. Return the JSON object directly in your response text.
 `;
 }

package/build/prompts/testGenerationPrompt.js

@@ -17,6 +17,7 @@ export function registerTestGenerationPrompt(mcpServer) {
 - ALWAYS SHOW STEPS TO GENERATE TEST USING MCP TOOLS AND NEVER SHOW THE CLI COMMANDS.
 - **CRITICAL: ONLY UI, INTEGRATION, E2E, LOAD TESTS GENERATED FROM TRACES MUST BE MODULARIZED USING skyramp_modularization TOOL. ADD A TASK TO MODULARIZE THE TEST USING skyramp_modularization TOOL AFTER GENERATING THESE(UI, INTEGRATION, E2E, LOAD) TESTS. DO NOT MODULARIZE THESE TESTS IF THEY ARE NOT GENERATED FROM TRACES.**
 - **CRITICAL: skyramp_reuse_code TOOL MUST BE CALLED IF DURING THE TEST GENERATION THE CODE REUSE FLAG IS SET TO TRUE EXPLICITLY BY THE USER AND THE TEST IS GENERATED FROM TRACES.**
+- **CRITICAL: DO NOT READ apiSchema FILES (OpenAPI/Swagger specifications). These files can be very large (often several MB). Simply pass the file path or URL to the test generation tool - the backend will handle reading and processing it. Never use file reading tools on apiSchema parameters.**
 
 **MANDATORY RULES**:
 1. **Priority Scores Must Remain Unchanged**: When a test type is missing required inputs (e.g., Playwright recordings, traces), **DO NOT**:

package/build/services/AnalyticsService.js

@@ -31,14 +31,6 @@ export class AnalyticsService {
             errorMessage: errorMessage,
             analyticsResult: params,
         });
-        // console.error(
-        //   "tool name ::: ",
-        //   toolName,
-        //   "params ::: ",
-        //   params,
-        //   "error message ::: ",
-        //   errorMessage,
-        // );
         await pushToolEvent(this.entryPoint, toolName, errorMessage, params);
     }
     /**

package/build/services/TestExecutionService.js

@@ -6,8 +6,9 @@ import { stripVTControlCharacters } from "util";
 import { logger } from "../utils/logger.js";
 const DEFAULT_TIMEOUT = 300000; // 5 minutes
 const MAX_CONCURRENT_EXECUTIONS = 5;
-const EXECUTOR_DOCKER_IMAGE = "skyramp/executor:v1.2.40";
+const EXECUTOR_DOCKER_IMAGE = "skyramp/executor:v1.3.0";
 const DOCKER_PLATFORM = "linux/amd64";
+const EXECUTION_PROGRESS_INTERVAL = 10000; // 10 seconds between progress updates during execution
 // Files and directories to exclude when mounting workspace to Docker container
 export const EXCLUDED_MOUNT_ITEMS = [
     "package-lock.json",
@@ -24,13 +25,13 @@ function findCommentStart(line) {
     let escaped = false;
     for (let i = 0; i < line.length; i++) {
         const char = line[i];
-        const nextChar = i + 1 < line.length ? line[i + 1] : '';
+        const nextChar = i + 1 < line.length ? line[i + 1] : "";
         // Handle escape sequences
         if (escaped) {
             escaped = false;
             continue;
         }
-        if (char === '\\') {
+        if (char === "\\") {
             escaped = true;
             continue;
         }
@@ -45,10 +46,10 @@ function findCommentStart(line) {
         }
         // Check for comments only if not in a string
         if (!inSingleQuote && !inDoubleQuote) {
-            if (char === '/' && nextChar === '/') {
+            if (char === "/" && nextChar === "/") {
                 return i;
             }
-            if (char === '#') {
+            if (char === "#") {
                 return i;
             }
         }
@@ -87,8 +88,8 @@ function filterComments(lines) {
             continue;
         }
         // Check for multi-line comment start/end (/* */)
-        if (trimmed.includes('/*')) {
-            if (trimmed.includes('*/')) {
+        if (trimmed.includes("/*")) {
+            if (trimmed.includes("*/")) {
                 // Single-line multi-line comment (e.g., /* comment */)
                 // Don't change state, just skip the line
                 continue;
@@ -100,13 +101,13 @@ function filterComments(lines) {
             }
         }
         if (inMultiLineComment) {
-            if (trimmed.includes('*/')) {
+            if (trimmed.includes("*/")) {
                 inMultiLineComment = false;
             }
             continue;
         }
         // Skip single-line comments
-        if (trimmed.startsWith('//') || trimmed.startsWith('#')) {
+        if (trimmed.startsWith("//") || trimmed.startsWith("#")) {
             continue;
         }
         // Remove inline comments from the line before processing
@@ -127,7 +128,7 @@ function filterComments(lines) {
 function detectSessionFiles(testFilePath) {
     try {
         const content = fs.readFileSync(testFilePath, "utf-8");
-        const lines = content.split('\n');
+        const lines = content.split("\n");
         const sessionFiles = [];
         // Pattern for TypeScript/JavaScript: storageState: '/path/to/file' or storageState: "/path/to/file"
         const tsJsPattern = /storageState:\s*['"]([^'"]+)['"]/g;
@@ -164,16 +165,17 @@ function detectSessionFiles(testFilePath) {
         return sessionFiles;
     }
     catch (error) {
-        logger.error(`Failed to detect session files in ${testFilePath}`, { error });
+        logger.error(`Failed to detect session files in ${testFilePath}`, {
+            error,
+        });
         return [];
     }
 }
 export class TestExecutionService {
     docker;
-    imageReady;
+    imageReady = null;
     constructor() {
         this.docker = new Docker();
-        this.imageReady = this.ensureDockerImage();
     }
     /**
      * Execute multiple tests in parallel batches
@@ -217,13 +219,42 @@ export class TestExecutionService {
     }
     /**
      * Execute a single test
+     * @param options Test execution options
+     * @param onProgress Optional callback for progress updates
      */
-    async executeTest(options) {
+    async executeTest(options, onProgress) {
         const startTime = Date.now();
         const executedAt = new Date().toISOString();
         logger.debug(`Executing test: ${options.testFile}`);
-        // Wait for Docker image to be ready
-        await this.imageReady;
+        // Check and ensure Docker image is ready with progress reporting
+        await onProgress?.({
+            phase: "docker-check",
+            message: "Checking Docker image availability...",
+            percent: 5,
+        });
+        const imageResult = await this.ensureDockerImage(onProgress);
+        if (imageResult.cached) {
+            await onProgress?.({
+                phase: "docker-check",
+                message: "Using cached Docker image",
+                percent: 20,
+                details: { cached: true },
+            });
+        }
+        else {
+            await onProgress?.({
+                phase: "docker-pull",
+                message: "Docker image pull completed",
+                percent: 20,
+                details: { cached: false },
+            });
+        }
+        // Report preparing phase
+        await onProgress?.({
+            phase: "preparing",
+            message: "Validating workspace and test file...",
+            percent: 25,
+        });
         // Validate workspace path - use accessSync for better validation
         const workspacePath = path.resolve(options.workspacePath);
         try {
@@ -240,6 +271,11 @@ export class TestExecutionService {
         if (!fs.existsSync(options.testFile)) {
             throw new Error(`Test file does not exist: ${options.testFile}`);
         }
+        await onProgress?.({
+            phase: "preparing",
+            message: "Preparing Docker container configuration...",
+            percent: 30,
+        });
         const containerMountPath = "/home/user";
         const dockerSocketPath = "/var/run/docker.sock";
         // Calculate relative test file path
@@ -330,8 +366,34 @@ export class TestExecutionService {
         }
         const stream = new DockerStream();
         try {
+            // Report executing phase
+            await onProgress?.({
+                phase: "executing",
+                message: "Starting test execution in Docker container...",
+                percent: 40,
+            });
             let statusCode = 0;
             let containerRef = null;
+            // Start periodic progress reporter during execution
+            // Runs in parallel with docker.run() to keep the AI agent informed
+            let progressIntervalHandle;
+            if (onProgress) {
+                let progressTick = 0;
+                progressIntervalHandle = setInterval(() => {
+                    progressTick++;
+                    const elapsed = Math.floor((Date.now() - startTime) / 1000);
+                    // Progress moves from 40% to 79% during execution phase
+                    // Slowly increment to show activity without reaching 80% (reserved for processing)
+                    const percent = Math.min(40 + progressTick * 2, 79);
+                    onProgress({
+                        phase: "executing",
+                        message: `Test running... (${elapsed}s elapsed)`,
+                        percent,
+                    }).catch(() => {
+                        // Ignore progress notification errors
+                    });
+                }, EXECUTION_PROGRESS_INTERVAL);
+            }
             // Run container with timeout
             const executionPromise = this.docker
                 .run(EXECUTOR_DOCKER_IMAGE, command, stream, {
@@ -367,14 +429,18 @@ export class TestExecutionService {
             });
             try {
                 statusCode = await Promise.race([executionPromise, timeoutPromise]);
-                // Clear the timeout timer if execution completes successfully
+                // Clear timers on successful completion
                 if (timeoutHandle)
                     clearTimeout(timeoutHandle);
+                if (progressIntervalHandle)
+                    clearInterval(progressIntervalHandle);
             }
             catch (error) {
-                // Clear the timeout timer on any error
+                // Clear all timers on error
                 if (timeoutHandle)
                     clearTimeout(timeoutHandle);
+                if (progressIntervalHandle)
+                    clearInterval(progressIntervalHandle);
                 // Cleanup container on timeout or other errors
                 if (containerRef !== null) {
                     const container = containerRef;
@@ -392,6 +458,12 @@ export class TestExecutionService {
                 }
                 throw error; // Re-throw the original error
             }
+            // Report processing phase
+            await onProgress?.({
+                phase: "processing",
+                message: "Test execution completed, processing results...",
+                percent: 80,
+            });
             const duration = Date.now() - startTime;
             const cleanOutput = stripVTControlCharacters(output);
             logger.info("Test execution completed", {
@@ -415,6 +487,14 @@ export class TestExecutionService {
                 output: cleanOutput,
                 exitCode: statusCode,
             };
+            // Report completion
+            await onProgress?.({
+                phase: "processing",
+                message: result.passed
+                    ? "Test passed successfully"
+                    : "Test execution completed with failures",
+                percent: 100,
+            });
             logger.debug(`Test ${result.passed ? "passed" : "failed"}: ${options.testFile}`);
             return result;
         }
@@ -436,17 +516,25 @@ export class TestExecutionService {
     }
     /**
      * Ensure Docker image is available
+     * @param onProgress Optional callback for progress updates during pull
+     * @returns Object indicating whether image was cached or pulled
      */
-    async ensureDockerImage() {
+    async ensureDockerImage(onProgress) {
         try {
             const images = await this.docker.listImages();
             const imageExists = images.some((img) => (img.RepoTags && img.RepoTags.includes(EXECUTOR_DOCKER_IMAGE)) ||
                 (img.RepoDigests && img.RepoDigests.includes(EXECUTOR_DOCKER_IMAGE)));
             if (imageExists) {
                 logger.debug(`Docker image ${EXECUTOR_DOCKER_IMAGE} already available`);
-                return;
+                return { cached: true };
             }
             logger.info(`Pulling Docker image ${EXECUTOR_DOCKER_IMAGE}...`);
+            await onProgress?.({
+                phase: "docker-pull",
+                message: `Pulling Docker image ${EXECUTOR_DOCKER_IMAGE}...`,
+                percent: 10,
+                details: { cached: false },
+            });
             await new Promise((resolve, reject) => {
                 this.docker.pull(EXECUTOR_DOCKER_IMAGE, { platform: DOCKER_PLATFORM }, (err, stream) => {
                     if (err)
@@ -458,9 +546,39 @@ export class TestExecutionService {
                             return reject(err);
                         logger.info(`Docker image ${EXECUTOR_DOCKER_IMAGE} pulled successfully`);
                         resolve(res);
+                    }, 
+                    // Progress callback for each layer
+                    (event) => {
+                        if (event.status && onProgress) {
+                            const progressMsg = event.progress || event.status;
+                            const layerId = event.id || "";
+                            // Calculate approximate progress (10-19% range during pull)
+                            let percent = 10;
+                            if (event.progressDetail?.current &&
+                                event.progressDetail?.total) {
+                                const layerProgress = event.progressDetail.current / event.progressDetail.total;
+                                percent = 10 + Math.floor(layerProgress * 9); // 10-19%
+                            }
+                            onProgress({
+                                phase: "docker-pull",
+                                message: `${event.status}${layerId ? ` [${layerId}]` : ""}: ${progressMsg}`,
+                                percent,
+                                details: {
+                                    cached: false,
+                                    pullProgress: {
+                                        current: event.progressDetail?.current,
+                                        total: event.progressDetail?.total,
+                                        layer: layerId,
+                                    },
+                                },
+                            }).catch(() => {
+                                // Ignore progress notification errors
+                            });
+                        }
                     });
                 });
             });
+            return { cached: false };
         }
         catch (error) {
             logger.error(`Failed to ensure Docker image: ${error.message}`);

package/build/tools/executeSkyrampTestTool.js

@@ -50,17 +50,42 @@ For detailed documentation visit: https://www.skyramp.dev/docs/quickstart`,
         _meta: {
             keywords: ["run test", "execute test"],
         },
-    }, async (params) => {
+    }, async (params, extra) => {
         let errorResult;
+        // Helper to send progress notifications to the MCP client
+        const sendProgress = async (progress, total, message) => {
+            const progressToken = extra._meta?.progressToken;
+            if (progressToken !== undefined) {
+                const notification = {
+                    method: "notifications/progress",
+                    params: {
+                        progressToken,
+                        progress,
+                        total,
+                        message,
+                    },
+                };
+                await extra.sendNotification(notification);
+            }
+        };
+        // Progress callback adapter for TestExecutionService
+        const onExecutionProgress = async (progress) => {
+            await sendProgress(progress.percent, 100, progress.message);
+        };
         try {
+            // Send initial progress
+            await sendProgress(0, 100, "Starting test execution...");
             const executionService = new TestExecutionService();
+            // Execute test with progress callback - reports Docker cache/pull status
             const result = await executionService.executeTest({
                 testFile: params.testFile,
                 workspacePath: params.workspacePath,
                 language: params.language,
                 testType: params.testType,
                 token: params.token,
-            });
+            }, onExecutionProgress);
+            // Progress is already reported by TestExecutionService
+            // Only report final status if not already at 100%
             if (!result.passed) {
                 errorResult = {
                     content: [
@@ -74,6 +99,7 @@ For detailed documentation visit: https://www.skyramp.dev/docs/quickstart`,
                 };
                 return errorResult;
             }
+            // Success - progress already reported by TestExecutionService
             return {
                 content: [
                     {

package/build/tools/generate-tests/generateContractRestTool.js

@@ -29,7 +29,9 @@ export function registerContractTestTool(server) {
     server.registerTool(TOOL_NAME, {
         description: `Generate a contract test using Skyramp's deterministic test generation platform.
 
-Contract tests ensure your API implementation matches its OpenAPI/Swagger specification exactly. They validate request/response schemas, status codes, headers, and data types to prevent contract violations and API breaking changes.`,
+Contract tests ensure your API implementation matches its OpenAPI/Swagger specification exactly. They validate request/response schemas, status codes, headers, and data types to prevent contract violations and API breaking changes.
+
+**IMPORTANT: If an apiSchema parameter (OpenAPI/Swagger file path or URL) is provided, DO NOT attempt to read or analyze the file contents. These files can be very large. Simply pass the path/URL to the tool - the backend will handle reading and processing the schema file.**`,
         inputSchema: contractTestSchema,
     }, async (params) => {
         const service = new ContractTestService();

package/build/tools/generate-tests/generateFuzzRestTool.js

@@ -25,7 +25,9 @@ export function registerFuzzTestTool(server) {
     server.registerTool(TOOL_NAME, {
         description: `Generate a fuzz test using Skyramp's deterministic test generation platform.
 
-Fuzz tests improve application security and reliability by sending invalid, malformed, or unexpected data to your API endpoints. They help discover edge cases, input validation issues, error handling problems, and potential security vulnerabilities.`,
+Fuzz tests improve application security and reliability by sending invalid, malformed, or unexpected data to your API endpoints. They help discover edge cases, input validation issues, error handling problems, and potential security vulnerabilities.
+
+**IMPORTANT: If an apiSchema parameter (OpenAPI/Swagger file path or URL) is provided, DO NOT attempt to read or analyze the file contents. These files can be very large. Simply pass the path/URL to the tool - the backend will handle reading and processing the schema file.**`,
         inputSchema: fuzzTestSchema,
         _meta: {
             keywords: ["negative test", "random test"],

package/build/tools/generate-tests/generateIntegrationRestTool.js

@@ -42,7 +42,9 @@ export function registerIntegrationTestTool(server) {
     server.registerTool(TOOL_NAME, {
         description: `Generate an integration test using Skyramp's deterministic test generation platform.
 
-Integration tests validate that multiple services, components, or modules work together correctly. They test complex user workflows, service interactions, data flow between systems, and ensure that integrated components function as expected in realistic scenarios.`,
+Integration tests validate that multiple services, components, or modules work together correctly. They test complex user workflows, service interactions, data flow between systems, and ensure that integrated components function as expected in realistic scenarios.
+
+**IMPORTANT: If an apiSchema parameter (OpenAPI/Swagger file path or URL) is provided, DO NOT attempt to read or analyze the file contents. These files can be very large. Simply pass the path/URL to the tool - the backend will handle reading and processing the schema file.**`,
         inputSchema: integrationTestSchema,
     }, async (params) => {
         const service = new IntegrationTestService();

package/build/tools/generate-tests/generateLoadRestTool.js

@@ -64,7 +64,9 @@ Load tests evaluate your application's performance, scalability, and stability u
 - IF THE USER DOES NOT PROVIDE LOAD PARAMETERS, THEN USE DEFAULT VALUES FOR LOAD TESTS:
   - loadDuration: "5" (5 seconds)
   - loadNumThreads: "1" (1 thread)
-  - Other load parameters should remain empty unless explicitly specified by the user`,
+  - Other load parameters should remain empty unless explicitly specified by the user
+
+**IMPORTANT: If an apiSchema parameter (OpenAPI/Swagger file path or URL) is provided, DO NOT attempt to read or analyze the file contents. These files can be very large. Simply pass the path/URL to the tool - the backend will handle reading and processing the schema file.**`,
         inputSchema: loadTestSchema,
         _meta: {
             keywords: ["load test", "performance test"],

package/build/tools/generate-tests/generateScenarioRestTool.js

@@ -8,7 +8,7 @@ const scenarioTestSchema = {
         .describe("Name of the test scenario with multiple steps. Describe the complete workflow you want to test, including all actions and their sequence.KEEP IT SHORT AND DESCRIPTIVE AS USED FOR FILE NAME"),
     destination: z
         .string()
-        .describe("Destination host (e.g., api.example.com) to be used for the test"),
+        .describe("Destination hostname or IP address for the test (e.g., api.example.com, localhost, 192.168.1.1). Do NOT include port numbers."),
     apiSchema: z
         .string()
         .describe("MUST be absolute path (/path/to/openapi.json) to the OpenAPI/Swagger schema file or a URL to the OpenAPI/Swagger schema file (e.g. https://demoshop.skyramp.dev/openapi.json). Required for accurate API mapping."),
@@ -77,6 +77,8 @@ The AI should parse the natural language scenario and provide:
 - AI-parsed HTTP method and path (required)
 - AI-parsed request/response bodies (optional)
 
+**IMPORTANT: If an apiSchema parameter (OpenAPI/Swagger file path or URL) is provided, DO NOT attempt to read or analyze the file contents. These files can be very large. Simply pass the path/URL to the tool - the backend will handle reading and processing the schema file.**
+
 **Note:** This tool generates one request at a time. Call multiple times for multi-step scenarios.`,
         inputSchema: scenarioTestSchema,
     }, async (params) => {

package/build/tools/generate-tests/generateSmokeRestTool.js

@@ -25,7 +25,9 @@ export function registerSmokeTestTool(server) {
     server.registerTool(TOOL_NAME, {
         description: `Generate a smoke test using Skyramp's deterministic test generation platform.
 
-Smoke testing is a preliminary check used to verify that an endpoint is accessible and returns a valid response. Smoke tests are useful for quickly identifying critical defects after significant changes. They provide rapid validation for basic functionality verification and endpoint accessibility testing.`,
+Smoke testing is a preliminary check used to verify that an endpoint is accessible and returns a valid response. Smoke tests are useful for quickly identifying critical defects after significant changes. They provide rapid validation for basic functionality verification and endpoint accessibility testing.
+
+**IMPORTANT: If an apiSchema parameter (OpenAPI/Swagger file path or URL) is provided, DO NOT attempt to read or analyze the file contents. These files can be very large. Simply pass the path/URL to the tool - the backend will handle reading and processing the schema file.**`,
         inputSchema: smokeTestSchema,
         _meta: {
             keywords: ["smoke test", "quick test"],

package/build/tools/generate-tests/generateUIRestTool.js

@@ -33,7 +33,8 @@ const uiTestSchema = {
     output: baseSchema.shape.output,
     outputDir: baseSchema.shape.outputDir,
     force: baseSchema.shape.force,
-    ...codeRefactoringSchema.shape,
+    codeReuse: codeRefactoringSchema.shape.codeReuse,
+    modularizeCode: codeRefactoringSchema.shape.modularizeCode.default(true),
 };
 export function registerUITestTool(server) {
     server.registerTool(TOOL_NAME, {

package/build/tools/test-recommendation/analyzeRepositoryTool.js

@@ -70,13 +70,18 @@ Output: Detailed RepositoryAnalysis JSON object with all repository characterist
                 repositoryPath: params.repositoryPath,
                 scanDepth: params.scanDepth,
             });
-            // Return prompt for LLM to execute
             const analysisPrompt = getRepositoryAnalysisPrompt(params.repositoryPath);
             return {
                 content: [
                     {
                         type: "text",
-                        text: `# Repository Analysis Request
+                        text: `⚠️ CRITICAL INSTRUCTION - READ FIRST:
+- Return the JSON object directly in your response text.
+- DO NOT create or save any files (no repository_analysis.json or .md files).
+- DO NOT use file write tools or save analysis to disk.
+- Output the complete analysis JSON inline in your response.
+
+# Repository Analysis Request
 
 Please analyze the repository at: \`${params.repositoryPath}\`
 

package/build/tools/test-recommendation/mapTestsTool.js

@@ -38,7 +38,7 @@ The scoring algorithm considers:
 - Available artifacts (OpenAPI specs, Playwright recordings)
 - Security requirements (authentication, sensitive data)
 
-Example usage:
+Example usage (direct JSON):
 \`\`\`
 {
   "analysisReport": "<RepositoryAnalysis JSON from skyramp_analyze_repository>",
@@ -50,7 +50,10 @@ Example usage:
 \`\`\`
 
 **CRITICAL RULES**:
-- DO NOT CREATE ANY .json or .md file during test mapping.
+- Return the JSON object directly in your response text.
+- DO NOT create or save any files (no test_mapping.json or .json/.md files).
+- DO NOT use file write tools or save analysis to disk.
+- Output the complete analysis JSON inline in your response.
 
 Output: TestMappingResult JSON with priority scores, feasibility, and reasoning for each test type.`,
         inputSchema: mapTestsSchema.shape,
@@ -140,7 +143,13 @@ Output: TestMappingResult JSON with priority scores, feasibility, and reasoning
                 summary: { highPriority, mediumPriority, lowPriority },
             };
             // Format output
-            const output = `# Test Priority Mapping
+            const output = `⚠️ CRITICAL INSTRUCTION - READ FIRST:
+- Return the JSON object directly in your response text.
+- DO NOT create or save any files (no repository_analysis.json or .md files).
+- DO NOT use file write tools or save analysis to disk.
+- Output the complete analysis JSON inline in your response.
+
+# Test Priority Mapping
 
 ## Summary
 - **High Priority**: ${mapping.summary.highPriority.join(", ") || "None"}

package/build/tools/test-recommendation/recommendTestsTool.js

@@ -107,6 +107,7 @@ ${prompt}
 2. Include guidance for missing artifacts with specific Skyramp tool commands
 3. Prioritize quick wins (tests with all artifacts available)
 4. Use actual endpoint paths and file paths from the repository analysis
+5. CRITICAL: Return the recommendations in text format, not JSON.
 
 PRE-REQUISITES:
 STEP 1: Call skyramp_analyze_repository to get repository analysis
@@ -117,6 +118,7 @@ STEP 2: Call skyramp_map_tests to get test priority scores and repository analys
 - DO NOT SHOW ANY PRIORITY BREAKDOWN IN THE OUTPUT.
 - DON'T MARK ANY TEST BLOCKED EVEN IF REQUIRED ARTIFACTS ARE MISSING.
 - DO NOT SHOW RESULTS IN .MD OR .JSON OR ANY OTHER FILE FORMAT.
+- DO NOT CREATE OR SAVE FILES - output everything in your response text.
 
 After analyzing the data above, return the complete JSON response following the structure defined in the prompt.
 **CRITICAL:**  At the end of the tool execution, MUST display the below message:

package/build/types/TestTypes.js

@@ -12,7 +12,13 @@ export var TestType;
 export const languageSchema = z.object({
     language: z
         .string()
-        .describe("Programming language for the generated test (default: python)"),
+        .refine((val) => {
+        const validLanguages = ["python", "typescript", "javascript", "java"];
+        return validLanguages.includes(val.toLowerCase());
+    }, {
+        message: "Language must be one of: python, typescript, javascript, java",
+    })
+        .describe("Programming language for the generated test (default: python). Must be one of: python, typescript, javascript, java"),
     framework: z
         .string()
         .describe("Testing framework to use (e.g., pytest for python, playwright for javascript and typescript, junit for java, etc.)"),
@@ -55,7 +61,15 @@ export const baseSchema = z.object({
     output: z
         .string()
         .optional()
-        .describe("Name of the output test file. Default value is empty string. If not provided, Skyramp will generate a default name for the test file."),
+        .refine((val) => {
+        if (!val || val === "")
+            return true; // Allow empty string or undefined
+        const validExtensions = [".py", ".ts", ".js", ".java"];
+        return validExtensions.some((ext) => val.endsWith(ext));
+    }, {
+        message: "Output file must have one of these extensions: .py, .ts, .js, .java",
+    })
+        .describe("Name of the output test file. Must have one of these extensions: .py, .ts, .js, .java. Default value is empty string. If not provided, Skyramp will generate a default name for the test file."),
     outputDir: z
         .string()
         .describe("MUST be absolute path to the directory where test files will be generated. If not provided, the CURRENT WORKING DIRECTORY will be used WITHOUT ANY SUBDIRECTORIES"),
@@ -104,7 +118,7 @@ export const baseTestSchema = {
     apiSchema: z
         .string()
         .default("")
-        .describe("MUST be absolute path(/path/to/openapi.json) to the OpenAPI/Swagger schema file or a URL to the OpenAPI/Swagger schema file(e.g. https://demoshop.skyramp.dev/openapi.json). DO NOT TRY TO ASSUME THE OPENAPI SCHEMA IF NOT PROVIDED"),
+        .describe("MUST be absolute path(/path/to/openapi.json) to the OpenAPI/Swagger schema file or a URL to the OpenAPI/Swagger schema file(e.g. https://demoshop.skyramp.dev/openapi.json). DO NOT TRY TO ASSUME THE OPENAPI SCHEMA IF NOT PROVIDED. NOTE TO AI ASSISTANTS: You do not need to read the contents of this file - simply pass the file path as the backend will read and process it."),
     pathParams: z
         .string()
         .default("")

package/build/utils/logger.js

@@ -15,8 +15,12 @@ export class McpLogger {
             message,
             ...(data && { data }),
         };
-        // Use console.error for structured logging (this goes to stderr)
-        console.error(JSON.stringify(logEntry));
+        try {
+            console.error(JSON.stringify(logEntry));
+        }
+        catch (error) {
+            // silently ignore
+        }
     }
     debug(message, data) {
         this.sendLogMessage("debug", message, data);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@skyramp/mcp",
-  "version": "0.0.46",
+  "version": "0.0.48",
   "main": "build/index.js",
   "type": "module",
   "bin": {
@@ -43,7 +43,7 @@
   },
   "dependencies": {
     "@modelcontextprotocol/sdk": "^1.24.3",
-    "@skyramp/skyramp": "^1.2.40",
+    "@skyramp/skyramp": "^1.3.0",
     "@playwright/test": "^1.55.0",
     "dockerode": "^4.0.6",
     "fast-glob": "^3.3.3",