@skyramp/mcp 0.4.3-rc.2 → 0.4.4-rc.1

package/build/recommendation/verifiers/issueTraceability.js

@@ -31,30 +31,6 @@ function quote(description) {
         .replace(/\s+/g, " ");
     return text.length > QUOTE_LENGTH ? `${text.slice(0, QUOTE_LENGTH)}…` : text;
 }
-/** The plan-time `defects:untested:<id>` objection, as the plan-time verifier ids it. */
-const UNTESTED_PREFIX = "defects:untested:";
-/** Defect ids whose `defects:untested:<id>` objection the agent closed when it
- *  registered the plan, by an answer or by a blocker. Both buckets: a blocker
- *  close is stored apart from an answer and is a close all the same. Read
- *  defensively: the plan comes off disk. */
-function answeredUntested(plan) {
-    const closed = [
-        ...(Array.isArray(plan?.answeredObjections) ? plan.answeredObjections : []),
-        ...(Array.isArray(plan?.unverifiedCloses) ? plan.unverifiedCloses : []),
-    ];
-    const ids = new Set();
-    for (const entry of closed) {
-        const objectionId = idKey(entry?.objection?.objectionId);
-        if (!objectionId.startsWith(UNTESTED_PREFIX))
-            continue;
-        if (idKey(entry?.answer).length === 0)
-            continue;
-        const id = idKey(objectionId.slice(UNTESTED_PREFIX.length));
-        if (id)
-            ids.add(id);
-    }
-    return ids;
-}
 /** The declared defect ids a `maintains` entry proves, as plan time judged them.
  *  Report time has no VerifyContext, so it reads plan time's own verdict on the
  *  entry's file, `existingCoverage:maintains:<file>`, rather than resolving the path
@@ -122,11 +98,7 @@ function locate(defect) {
  *  accepted that entry — the same claim a planned test makes. See
  *  {@link maintainedDefects} for how that acceptance is read.
  *  `deliveredMatchesPlan:maintains:` checks separately that the edit landed.
- *
- *  A declared `defectId` also counts when the agent answered `defects:untested:<id>`
- *  at plan time (or closed it with a blocker). The plan-time verifier asked why no
- *  test proves that defect and the agent said; asking again here is the same
- *  question twice, and the answer is already published with the plan. */
+ */
 export function checkIssueTraceability(plan, delivered, issues, lane = NORMAL_LANE) {
     const shipped = new Set(delivered.map((test) => idKey(test?.plannedTestId)).filter(Boolean));
     const plannedTests = Array.isArray(plan?.plannedTests)
@@ -173,7 +145,6 @@ export function checkIssueTraceability(plan, delivered, issues, lane = NORMAL_LA
             bucket.set(defectId, [...(bucket.get(defectId) ?? []), id]);
         }
     }
-    const answered = answeredUntested(plan);
     const maintained = maintainedDefects(plan);
     const objections = [];
     issues.forEach((issue, index) => {
@@ -187,9 +158,7 @@ export function checkIssueTraceability(plan, delivered, issues, lane = NORMAL_LA
         const defectId = idKey(issue?.defectId);
         const declaredDefect = defectId.length > 0 && known.has(defectId);
         if (declaredDefect &&
-            ((provenBy.get(defectId) ?? []).length > 0 ||
-                maintained.has(defectId) ||
-                answered.has(defectId))) {
+            ((provenBy.get(defectId) ?? []).length > 0 || maintained.has(defectId))) {
             return;
         }
         const named = [];
@@ -204,7 +173,7 @@ export function checkIssueTraceability(plan, delivered, issues, lane = NORMAL_LA
                 named.push(`defectId "${defectId}" is cited by no ${word} test that expects to fail: ${(citedWeakly.get(defectId) ?? []).join(", ")} cite it and the plan declares each as expecting to pass`);
             }
             else {
-                named.push(`defectId "${defectId}" is cited by no ${word} test that expects to fail, no accepted \`maintains\` entry proves it — an entry naming a file that is not in the checkout is refused — and no plan-time answer closed ${UNTESTED_PREFIX}${defectId}`);
+                named.push(`defectId "${defectId}" is cited by no ${word} test that expects to fail, no accepted \`maintains\` entry proves it — an entry naming a file that is not in the checkout is refused`);
             }
         }
         const where = String(issue?.sourceFile ?? "").trim() || "no sourceFile";
@@ -217,7 +186,7 @@ export function checkIssueTraceability(plan, delivered, issues, lane = NORMAL_LA
             verifier: "issueTraceability",
             message: `No test proves this bug: "${quote(issue?.description)}" (${where}).`,
             evidence: `issuesFound[${index}]: ${named.length > 0 ? named.join("; ") : "no plannedTestId and no defectId"}.`,
-            suggestion: "Set `plannedTestId` to the delivered test that proves this bug — one the plan declares with `expected.outcome: fail`, so it stays red until the bug is fixed — or set `defectId` to a declared plan defect such a test cites. If no test proves it, answer in one line why.",
+            suggestion: "Set `plannedTestId` to the delivered test that proves this bug — one the plan declares with `expected.outcome: fail`, so it stays red until the bug is fixed — or set `defectId` to a declared plan defect such a test cites. If something stopped the test — a service that is not running, the one credential the run holds — answer and set `blocker` to it. A reason the bug is not worth testing is not a blocker.",
         });
     });
     return objections;

package/build/tools/executeSkyrampTestTool.js

@@ -17,6 +17,7 @@ import { StateManager, currentRunStateFile, getPrimaryRepository, getTestsRepoDi
 import { DriftAction, } from "../types/TestAnalysis.js";
 import { logger } from "../utils/logger.js";
 import { toolError } from "../utils/utils.js";
+import { staleSdkRefusal } from "../utils/repoSdkVersion.js";
 import { recordExecutionVideo } from "./execution-video-state.js";
 import { stageGeneratedPaths } from "../utils/gitStaging.js";
 import { walkDir } from "../utils/fileWalk.js";
@@ -1142,6 +1143,13 @@ export function registerExecuteSkyrampTestTool(server) {
                         errorResult = toolError(`commandOverride does not contain ${testFileNames(params.testFile).join(" or ")}, the name of testFile. The command was not run.`);
                         return errorResult;
                     }
+                    // SKYR-4548: only a host run loads the repository's own SDK; the
+                    // executor container supplies its own whatever the repository pins.
+                    const staleSdk = staleSdkRefusal(params.testFile, params.language);
+                    if (staleSdk) {
+                        errorResult = toolError(staleSdk);
+                        return errorResult;
+                    }
                 }
                 // Hashed once, before any state write: recorded with the reservation
                 // so the NEXT run can tell an unchanged re-run from a fixed file.

package/build/tools/generate-tests/generateE2ERestTool.d.ts

@@ -7,6 +7,7 @@ export declare class E2ETestService extends TestGenerationService {
     protected buildGenerationOptions(params: BaseTestParams & {
         responseData?: string;
         playwrightInput?: string;
+        testName?: string;
     }): any;
     protected handleApiAnalysis(_params: BaseTestParams): Promise<CallToolResult | null>;
 }

package/build/tools/generate-tests/generateE2ERestTool.js

@@ -1,5 +1,5 @@
 import { z } from "zod";
-import { baseSchema, baseTraceSchema, TestType, codeRefactoringSchema, } from "../../types/TestTypes.js";
+import { baseSchema, baseTraceSchema, TestType, codeRefactoringSchema, testNameField, } from "../../types/TestTypes.js";
 import { TestGenerationService, } from "../../services/TestGenerationService.js";
 import { AnalyticsService } from "../../services/AnalyticsService.js";
 import { toolError } from "../../utils/utils.js";
@@ -16,6 +16,7 @@ const e2eTestSchema = {
         "from a Playwright zip that already produced one is rejected — one recording covers one scenario, and reusing " +
         "it manufactures identical tests under names the recording never covered. Set true only when the same " +
         "recording genuinely covers the new scenario end-to-end."),
+    testName: testNameField("testE2e"),
     ...codeRefactoringSchema.shape,
     ...baseSchema.shape,
 };
@@ -28,6 +29,11 @@ export class E2ETestService extends TestGenerationService {
             ...super.buildBaseGenerationOptions(params),
             responseData: params.responseData,
             playwrightInput: params.playwrightInput,
+            // SKYR-4468: free-form title for the generated test, so the spec names the
+            // behaviour it verifies instead of the generic `testE2e`. The option is
+            // inert until the pinned @skyramp/skyramp release carries it — the client
+            // ignores options it does not know, so passing it early is safe.
+            testName: params.testName,
         };
     }
     async handleApiAnalysis(_params) {
@@ -44,6 +50,8 @@ End-to-End tests validate complete user journeys by testing the entire applicati
 TRACE & UI INTEGRATION:
 E2E tests require both trace files (capturing backend API interactions) and Playwright recordings (capturing UI interactions captured using start_trace_collection tool) to generate comprehensive tests that validate the complete user experience.
 
+**Test title:** Pass \`testName\` with the behaviour this recording verifies (e.g. "Add to cart from search results"), taken from the recommendation the recording came from. Omit it and the spec is titled \`testE2e\`, which claims nothing.
+
 **CRITICAL - When using trace parameter:**
 If \`trace\` parameter is provided (path to a trace file), DO NOT pass \`apiSchema\` or \`endpointURL\` parameters. The trace file already contains all necessary endpoint and schema information. Passing both will cause test generation to fail.`,
         inputSchema: e2eTestSchema,

package/build/tools/generate-tests/generateUIRestTool.d.ts

@@ -8,6 +8,7 @@ export declare class UITestService extends TestGenerationService {
     protected buildGenerationOptions(params: BaseTestParams & {
         playwrightInput?: string;
         uniqueUIFormInputs?: string[];
+        testName?: string;
     }): any;
     protected handleApiAnalysis(_params: BaseTestParams): Promise<CallToolResult | null>;
     generateTest(params: BaseTestParams & Record<string, any>): Promise<CallToolResult>;
@@ -19,6 +20,7 @@ export declare const uiTestSchema: {
     include: z.ZodOptional<z.ZodDefault<z.ZodArray<z.ZodString, "many">>>;
     exclude: z.ZodOptional<z.ZodDefault<z.ZodArray<z.ZodString, "many">>>;
     uniqueUIFormInputs: z.ZodOptional<z.ZodArray<z.ZodString, "many">>;
+    testName: z.ZodOptional<z.ZodString>;
     output: z.ZodEffects<z.ZodOptional<z.ZodString>, string | undefined, string | undefined>;
     outputDir: z.ZodString;
     force: z.ZodDefault<z.ZodBoolean>;

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

@@ -1,4 +1,4 @@
-import { baseSchema, languageSchema, TestType, codeRefactoringSchema, baseTraceSchema, } from "../../types/TestTypes.js";
+import { baseSchema, languageSchema, TestType, codeRefactoringSchema, baseTraceSchema, testNameField, } from "../../types/TestTypes.js";
 import { z } from "zod";
 import { AnalyticsService } from "../../services/AnalyticsService.js";
 import { TestGenerationService, } from "../../services/TestGenerationService.js";
@@ -21,6 +21,11 @@ export class UITestService extends TestGenerationService {
             // SKYR-4012: create-form inputs the agent flagged as unique-constrained,
             // so Skyramp codegen types a per-run-unique value into them.
             uniqueUIFormInputs: params.uniqueUIFormInputs,
+            // SKYR-4468: free-form title for the generated test, so the spec names the
+            // behaviour it verifies instead of the generic `testUi`. The option is
+            // inert until the pinned @skyramp/skyramp release carries it — the client
+            // ignores options it does not know, so passing it early is safe.
+            testName: params.testName,
         };
     }
     async handleApiAnalysis(_params) {
@@ -86,6 +91,7 @@ export const uiTestSchema = {
         .array(z.string().min(1))
         .optional()
         .describe("Create-form inputs whose value is UNIQUE-constrained (e.g. a name/slug that the backend rejects as a duplicate). Identify each by the accessible label or selector you typed into during recording (the same value passed to getByLabel/getByPlaceholder/getByRole, or a raw selector). Skyramp types a per-run-unique value into these inputs — and reuses it in later assertions — so the generated UI test can be re-run without hitting a duplicate/conflict error. Inspect the app's data model for unique constraints; omit when no create-form field is unique-constrained."),
+    testName: testNameField("testUi"),
     output: baseSchema.shape.output,
     outputDir: baseSchema.shape.outputDir,
     force: baseSchema.shape.force,
@@ -123,6 +129,8 @@ This tells you exactly which frontend files changed so you record traces for the
 
 **Idempotency (re-runnable create flows):** If the recorded flow creates a resource by typing into a UNIQUE-constrained field (a name/slug the backend rejects as a duplicate on a second run), pass those inputs in \`uniqueUIFormInputs\`. Skyramp then types a per-run-unique value into them (and reuses it in later assertions) so the generated test can be re-run without a duplicate/conflict failure.
 
+**Test title:** Pass \`testName\` with the behaviour this recording verifies (e.g. "Add to cart from search results"), taken from the recommendation the recording came from. Omit it and the spec is titled \`testUi\`, which claims nothing.
+
 **DOM Analyzer tools for blueprint-aware recording:** alongside the basic \`browser_*\` interaction tools, the Skyramp MCP exposes \`browser_blueprint\` (canonical PageBlueprint capture). This enables semantic target selection and delta-derived assertions: capture a blueprint before each meaningful action, perform the action, then capture again — the delta between the two grounds your assertions in observable state changes rather than author guesses about what "success" looks like.
 
 **CRITICAL: Do NOT use skyramp_start_trace_collection/skyramp_stop_trace_collection for UI test recording — use browser_* tools + skyramp_export_zip instead.**`,

package/build/tools/submitReportTool.d.ts

@@ -25,9 +25,9 @@ export declare const targetElementSchema: z.ZodObject<{
     widgetType: z.ZodOptional<z.ZodEnum<["native", "custom", "unknown"]>>;
 }, "strip", z.ZodTypeAny, {
     role: string;
-    accessibleName: string;
     testId: string | null;
     stableId: string | null;
+    accessibleName: string;
     contextText: string[] | null;
     mutability?: "unknown" | "mutable" | "immutable" | undefined;
     widgetType?: "unknown" | "custom" | "native" | undefined;
@@ -73,9 +73,9 @@ export declare const newTestSchema: z.ZodEffects<z.ZodEffects<z.ZodObject<{
         widgetType: z.ZodOptional<z.ZodEnum<["native", "custom", "unknown"]>>;
     }, "strip", z.ZodTypeAny, {
         role: string;
-        accessibleName: string;
         testId: string | null;
         stableId: string | null;
+        accessibleName: string;
         contextText: string[] | null;
         mutability?: "unknown" | "mutable" | "immutable" | undefined;
         widgetType?: "unknown" | "custom" | "native" | undefined;
@@ -112,9 +112,9 @@ export declare const newTestSchema: z.ZodEffects<z.ZodEffects<z.ZodObject<{
     scenarioFile?: string | undefined;
     targetElements?: {
         role: string;
-        accessibleName: string;
         testId: string | null;
         stableId: string | null;
+        accessibleName: string;
         contextText: string[] | null;
         mutability?: "unknown" | "mutable" | "immutable" | undefined;
         widgetType?: "unknown" | "custom" | "native" | undefined;
@@ -164,9 +164,9 @@ export declare const newTestSchema: z.ZodEffects<z.ZodEffects<z.ZodObject<{
     scenarioFile?: string | undefined;
     targetElements?: {
         role: string;
-        accessibleName: string;
         testId: string | null;
         stableId: string | null;
+        accessibleName: string;
         contextText: string[] | null;
         mutability?: "unknown" | "mutable" | "immutable" | undefined;
         widgetType?: "unknown" | "custom" | "native" | undefined;
@@ -216,9 +216,9 @@ export declare const newTestSchema: z.ZodEffects<z.ZodEffects<z.ZodObject<{
     scenarioFile?: string | undefined;
     targetElements?: {
         role: string;
-        accessibleName: string;
         testId: string | null;
         stableId: string | null;
+        accessibleName: string;
         contextText: string[] | null;
         mutability?: "unknown" | "mutable" | "immutable" | undefined;
         widgetType?: "unknown" | "custom" | "native" | undefined;
@@ -274,9 +274,9 @@ export declare const additionalRecommendationSchema: z.ZodEffects<z.ZodEffects<z
         widgetType: z.ZodOptional<z.ZodEnum<["native", "custom", "unknown"]>>;
     }, "strip", z.ZodTypeAny, {
         role: string;
-        accessibleName: string;
         testId: string | null;
         stableId: string | null;
+        accessibleName: string;
         contextText: string[] | null;
         mutability?: "unknown" | "mutable" | "immutable" | undefined;
         widgetType?: "unknown" | "custom" | "native" | undefined;
@@ -318,9 +318,9 @@ export declare const additionalRecommendationSchema: z.ZodEffects<z.ZodEffects<z
     scenarioName?: string | undefined;
     targetElements?: {
         role: string;
-        accessibleName: string;
         testId: string | null;
         stableId: string | null;
+        accessibleName: string;
         contextText: string[] | null;
         mutability?: "unknown" | "mutable" | "immutable" | undefined;
         widgetType?: "unknown" | "custom" | "native" | undefined;
@@ -386,9 +386,9 @@ export declare const additionalRecommendationSchema: z.ZodEffects<z.ZodEffects<z
     scenarioName?: string | undefined;
     targetElements?: {
         role: string;
-        accessibleName: string;
         testId: string | null;
         stableId: string | null;
+        accessibleName: string;
         contextText: string[] | null;
         mutability?: "unknown" | "mutable" | "immutable" | undefined;
         widgetType?: "unknown" | "custom" | "native" | undefined;
@@ -454,9 +454,9 @@ export declare const additionalRecommendationSchema: z.ZodEffects<z.ZodEffects<z
     scenarioName?: string | undefined;
     targetElements?: {
         role: string;
-        accessibleName: string;
         testId: string | null;
         stableId: string | null;
+        accessibleName: string;
         contextText: string[] | null;
         mutability?: "unknown" | "mutable" | "immutable" | undefined;
         widgetType?: "unknown" | "custom" | "native" | undefined;