@muggleai/works 5.31.0-staging.268 → 5.31.0-staging.270

package/.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
       "name": "muggleai",
       "source": "./plugin",
       "description": "Real-browser E2E acceptance for your coding agent: test the flows you changed and open the PR with screenshots. Built for flows other tools can't reach: managed test identities with live inboxes (magic links, email OTP and 2FA, verification mail, password resets), an encrypted secret store for login and test credentials, and a test graph that runs prerequisite flows first. Plain-English tests become replayable scripts on localhost or preview URLs.",
-      "version": "5.31.0-staging.268"
+      "version": "5.31.0-staging.270"
     }
   ]
 }

package/README.md

@@ -273,7 +273,7 @@ so the result carries a viewUrl straight to the dashboard
 
 ## MCP tool reference
 
-108 tools across authentication, projects, use cases, test cases, scripts, workflows, local execution, reporting, secrets, billing, and administration. Call them directly from any MCP client to build your own pipeline.
+125 tools across authentication, projects, use cases, test cases, test states, scripts, workflows, local execution, reporting, secrets, billing, and administration. Call them directly from any MCP client to build your own pipeline.
 
 **Authentication** — `muggle-remote-auth-status`, `-login`, `-poll`, `-logout`, plus `-api-key-create`, `-list`, `-get`, `-revoke`.
 
@@ -301,6 +301,13 @@ so the result carries a viewUrl straight to the dashboard
 | `muggle-remote-test-case-bulk-preview-submit` | Async batch preview (~50% cheaper) |
 | `muggle-remote-test-plan-graph-rebuild` | Rebuild the prerequisite graph |
 
+| Test states | Purpose |
+| :---------- | :------ |
+| `muggle-remote-test-chaining-catalog-get` / `-settings-update` | Read a project's states and roles; set its fresh account set state and parallel account set limit |
+| `muggle-remote-test-state-create` / `-update` / `-delete` | Manage states: facts about one account set's data, plus a check |
+| `muggle-remote-test-role-create` / `-update` / `-delete` | Manage roles: created or supplied accounts a test acts as |
+| `muggle-remote-test-chaining-plan-get` | Preview the account set plan without running anything |
+
 Bulk-preview submissions return a `jobId` immediately. Poll `muggle-remote-bulk-preview-job-get` until terminal, then persist with the matching `-create` tool. `-list` and `-cancel` round out the set.
 
 | Scripts and workflows | Purpose |

package/dist/{chunk-J2DEXYH7.js → chunk-W3HJBL5U.js}

@@ -1,4 +1,4 @@
-import { __export, getLogger, getConfig, createChildLogger, buildElectronAppReleaseAssetUrl, getActiveRuntimeTarget, getAuthService, hasApiKey, getElectronAppVersion, getElectronAppDir, getPlatformKey, reconcileProjectPreferences, getDataDir2, PREFERENCES_FILE_NAME, isFirstRun, writePreferences, DEFAULT_PREFERENCES, isElectronAppInstalled, getElectronAppChecksums, getChecksumForPlatform, getElectronAppSignedFromVersion, getReleaseSignerIdentityUri, verifyFileChecksum, resolveElectronAppExecutablePath, calculateFileChecksum, initTelemetry, Surface, ServiceName, track, EventName, getQaTools, getLocalQaTools, performLogout, assertDeviceCodeClientProvisioned, performLogin, getActiveElectronAppReleaseStream, buildToolInputJsonSchema, toolRequiresAuth, getCallerCredentials, parseElectronAppDirName, getElectronAppReleaseTagPrefix, hasShownDisclosure, markDisclosureShown, getBundledElectronAppVersion, getElectronAppVersionSource, getCredentialsFilePath, ElectronAppReleaseStream, buildElectronAppChecksumsUrl, __require, getDisclosureCopy, buildOnboardingPlan, applyOnboardingAnswers, needsOnboarding, buildLlmSourcePlan, describeLlmSourceStatus, resolveLlmSourceStatus, readRunView, RunViewError, renderRunViewText, writeRunViewHtml, openBrowserUrl, evaluateCheckoutFreshness, runGitCommand, BLOCKING_FRESHNESS_VERDICTS, claimServiceStack, BUSY_EXIT_CODE, checkServiceStack, releaseServiceStack, readServiceStackStatus, resolvePreferences, OnboardingBlanketChoice, PreferenceKey, LlmSource, applyLlmSourceSelection, describeStackMismatch, getDataDir, DEFAULT_LOCAL_LLM_BASE_URL, isValidLocalBaseUrl, local_exports } from './chunk-YLQSNSSM.js';
+import { __export, getLogger, getConfig, createChildLogger, buildElectronAppReleaseAssetUrl, getActiveRuntimeTarget, getAuthService, hasApiKey, getElectronAppVersion, getElectronAppDir, getPlatformKey, reconcileProjectPreferences, getDataDir2, PREFERENCES_FILE_NAME, isFirstRun, writePreferences, DEFAULT_PREFERENCES, isElectronAppInstalled, getElectronAppChecksums, getChecksumForPlatform, getElectronAppSignedFromVersion, getReleaseSignerIdentityUri, verifyFileChecksum, resolveElectronAppExecutablePath, calculateFileChecksum, initTelemetry, Surface, ServiceName, track, EventName, getQaTools, getLocalQaTools, performLogout, assertDeviceCodeClientProvisioned, performLogin, getActiveElectronAppReleaseStream, buildToolInputJsonSchema, toolRequiresAuth, getCallerCredentials, parseElectronAppDirName, getElectronAppReleaseTagPrefix, hasShownDisclosure, markDisclosureShown, getBundledElectronAppVersion, getElectronAppVersionSource, getCredentialsFilePath, ElectronAppReleaseStream, buildElectronAppChecksumsUrl, __require, getDisclosureCopy, buildOnboardingPlan, applyOnboardingAnswers, needsOnboarding, buildLlmSourcePlan, describeLlmSourceStatus, resolveLlmSourceStatus, readRunView, RunViewError, renderRunViewText, writeRunViewHtml, openBrowserUrl, evaluateCheckoutFreshness, runGitCommand, BLOCKING_FRESHNESS_VERDICTS, claimServiceStack, BUSY_EXIT_CODE, checkServiceStack, releaseServiceStack, readServiceStackStatus, resolvePreferences, OnboardingBlanketChoice, PreferenceKey, LlmSource, applyLlmSourceSelection, describeStackMismatch, getDataDir, DEFAULT_LOCAL_LLM_BASE_URL, isValidLocalBaseUrl, local_exports } from './chunk-XUANXZ2O.js';
 import { Server } from '@modelcontextprotocol/sdk/server/index.js';
 import { ListToolsRequestSchema, CallToolRequestSchema, ListResourcesRequestSchema, ReadResourceRequestSchema } from '@modelcontextprotocol/sdk/types.js';
 import { v4 } from 'uuid';
@@ -1165,7 +1165,7 @@ async function resolveGsScreenshotUrls(report, opts) {
     if (gsUrls.length === 0) {
       return report;
     }
-    const mcps = await import('./src-7OY7H63R.js');
+    const mcps = await import('./src-J6AIYTLA.js');
     const credentials = await mcps.getCallerCredentialsAsync();
     if (!credentials.bearerToken && !credentials.apiKey) {
       stderrWrite(

package/dist/{chunk-YLQSNSSM.js → chunk-XUANXZ2O.js}

@@ -6375,6 +6375,9 @@ var PinnedAccountSchema = z.object({
 
 // packages/mcps/src/mcp/e2e/contracts/constants.ts
 var WORKFLOW_RUNTIME_ID_NOT_RUN_ID_HINT = "Use the `workflowRuntimeId` returned when the workflow was started \u2014 not a run `id` or a test runs summary's `latestWorkflowRunId`, which are run ids and are not found here.";
+var TEST_STATE_PARALLEL_ACCOUNT_SET_LIMIT_MIN = 1;
+var TEST_STATE_PARALLEL_ACCOUNT_SET_LIMIT_MAX = 20;
+var TEST_STATE_PARALLEL_ACCOUNT_SET_LIMIT_DEFAULT = 5;
 var FanOutChoice = /* @__PURE__ */ ((FanOutChoice2) => {
   FanOutChoice2["FollowPolicy"] = "followPolicy";
   FanOutChoice2["PrimaryOnly"] = "primaryOnly";
@@ -6401,6 +6404,85 @@ var SecretFieldEnum = /* @__PURE__ */ ((SecretFieldEnum2) => {
   SecretFieldEnum2["BILLING_POSTAL_CODE"] = "billing_postal_code";
   return SecretFieldEnum2;
 })(SecretFieldEnum || {});
+
+// packages/mcps/src/mcp/e2e/contracts/test-state-types.ts
+var TestStateFactScopeEnum = /* @__PURE__ */ ((TestStateFactScopeEnum2) => {
+  TestStateFactScopeEnum2["ACCOUNT_SET"] = "ACCOUNT_SET";
+  TestStateFactScopeEnum2["ACCOUNT"] = "ACCOUNT";
+  return TestStateFactScopeEnum2;
+})(TestStateFactScopeEnum || {});
+var TestRoleSourceEnum = /* @__PURE__ */ ((TestRoleSourceEnum2) => {
+  TestRoleSourceEnum2["CREATED"] = "CREATED";
+  TestRoleSourceEnum2["SUPPLIED"] = "SUPPLIED";
+  return TestRoleSourceEnum2;
+})(TestRoleSourceEnum || {});
+
+// packages/mcps/src/mcp/e2e/contracts/test-state-schemas.ts
+var TestStateProjectIdSchema = MuggleEntityIdSchema.describe("Project ID (UUID) that owns the test states and roles");
+var TestStateIdSchema = MuggleEntityIdSchema.describe("Test state ID (UUID)");
+var TestRoleIdSchema = MuggleEntityIdSchema.describe("Test role ID (UUID)");
+var TestStateFactSchema = z.object({
+  statement: z.string().min(1).describe("What holds, e.g. 'the thread shows one unread message'"),
+  scope: z.enum(TestStateFactScopeEnum).describe(
+    "ACCOUNT_SET: only this account set's data, e.g. the patient's thread with the doctor. ACCOUNT: one account as a whole, e.g. the doctor's unread count"
+  ),
+  roleId: MuggleEntityIdSchema.nullable().describe("Test role ID (UUID) whose account an ACCOUNT fact is about; null for an ACCOUNT_SET fact")
+});
+var TestStateFactsSchema = z.array(TestStateFactSchema).describe("Every fact the state's tests rely on");
+var TestStateCheckSchema = z.string().describe("Quick read-only check that an account set is in this state, e.g. 'the thread shows one unread message'; may be empty");
+var TestRoleSourceSchema = z.enum(TestRoleSourceEnum).describe("CREATED: Muggle signs the account up itself. SUPPLIED: the customer provides the account");
+var TestRoleAllowsChangesSchema = z.boolean().describe("SUPPLIED only: the customer allows tests to change the account's account-wide facts when a proven way back restores them");
+var TestChainingCatalogGetInputSchema = z.object({
+  projectId: TestStateProjectIdSchema
+});
+var TestChainingSettingsUpdateInputSchema = z.object({
+  projectId: TestStateProjectIdSchema,
+  freshAccountSetStateId: MuggleEntityIdSchema.nullable().optional().describe("Test state ID (UUID) a newly formed account set starts in, e.g. 'New patient'; null clears it"),
+  parallelAccountSetLimit: z.number().int().min(TEST_STATE_PARALLEL_ACCOUNT_SET_LIMIT_MIN).max(TEST_STATE_PARALLEL_ACCOUNT_SET_LIMIT_MAX).optional().describe(`Most account sets a batch runs at once (default ${TEST_STATE_PARALLEL_ACCOUNT_SET_LIMIT_DEFAULT})`)
+});
+var TestStateCreateInputSchema = z.object({
+  projectId: TestStateProjectIdSchema,
+  name: z.string().min(1).describe("State name, unique per project ignoring case, e.g. 'Message unread'"),
+  facts: TestStateFactsSchema.optional(),
+  check: TestStateCheckSchema.optional()
+});
+var TestStateUpdateInputSchema = z.object({
+  projectId: TestStateProjectIdSchema,
+  stateId: TestStateIdSchema,
+  name: z.string().min(1).optional().describe("New state name, unique per project ignoring case"),
+  facts: TestStateFactsSchema.optional(),
+  check: TestStateCheckSchema.optional()
+});
+var TestStateDeleteInputSchema = z.object({
+  projectId: TestStateProjectIdSchema,
+  stateId: TestStateIdSchema
+});
+var TestRoleCreateInputSchema = z.object({
+  projectId: TestStateProjectIdSchema,
+  name: z.string().min(1).describe("Role name, unique per project ignoring case, e.g. 'Patient'"),
+  source: TestRoleSourceSchema,
+  allowsChanges: TestRoleAllowsChangesSchema.optional()
+});
+var TestRoleUpdateInputSchema = z.object({
+  projectId: TestStateProjectIdSchema,
+  roleId: TestRoleIdSchema,
+  name: z.string().min(1).optional().describe("New role name, unique per project ignoring case"),
+  source: TestRoleSourceSchema.optional(),
+  allowsChanges: TestRoleAllowsChangesSchema.optional()
+});
+var TestRoleDeleteInputSchema = z.object({
+  projectId: TestStateProjectIdSchema,
+  roleId: TestRoleIdSchema
+});
+var TestChainingPlanGetInputSchema = z.object({
+  projectId: TestStateProjectIdSchema,
+  isProduction: z.boolean().optional().describe("True to plan for production, which runs at the floor with no extra account sets for speed (default false)")
+});
+var TestCaseTestStateFieldsSchema = z.object({
+  actsAsRoleIds: z.array(MuggleEntityIdSchema).nullable().optional().describe("Test role IDs (UUID) the test acts as, in order; null clears them"),
+  prerequisiteStateId: MuggleEntityIdSchema.nullable().optional().describe("Test state ID (UUID) the test needs to start; null clears it"),
+  resultStateId: MuggleEntityIdSchema.nullable().optional().describe("Test state ID (UUID) the test leaves when it passes; the prerequisite's ID for an observation; null clears it")
+});
 var FanOutPlanInputSchema = z.object({
   projectId: z.string().min(1).describe("Project ID"),
   testCaseId: z.string().min(1).describe("Test case ID whose fan-out to plan"),
@@ -6720,7 +6802,8 @@ var TestCaseCreateInputSchema = z.object({
   category: z.string().optional().describe("Test case category"),
   automated: z.boolean().optional().describe("Whether this test case is automated (default: true)"),
   pinnedAccount: PinnedAccountSchema.optional().describe("Project secrets the test case signs in as. Omit to leave it unpinned"),
-  alwaysStartsFreshSession: z.boolean().optional().describe(ALWAYS_STARTS_FRESH_SESSION_DESCRIPTION)
+  alwaysStartsFreshSession: z.boolean().optional().describe(ALWAYS_STARTS_FRESH_SESSION_DESCRIPTION),
+  ...TestCaseTestStateFieldsSchema.shape
 });
 var TestCaseUpdateInputSchema = z.object({
   testCaseId: IdSchema.describe("Test case ID (UUID) to update"),
@@ -6736,7 +6819,8 @@ var TestCaseUpdateInputSchema = z.object({
   category: z.string().optional().describe("Updated category"),
   automated: z.boolean().optional().describe("Whether this test case is automated"),
   pinnedAccount: PinnedAccountSchema.nullable().optional().describe("Project secrets the test case signs in as; null clears the pin"),
-  alwaysStartsFreshSession: z.boolean().optional().describe(ALWAYS_STARTS_FRESH_SESSION_DESCRIPTION)
+  alwaysStartsFreshSession: z.boolean().optional().describe(ALWAYS_STARTS_FRESH_SESSION_DESCRIPTION),
+  ...TestCaseTestStateFieldsSchema.shape
 });
 var TestScriptListInputSchema = z.object({
   projectId: IdSchema.describe("Project ID (UUID) to list test scripts for"),
@@ -7611,6 +7695,32 @@ function mapTestScriptsSummary(response) {
   };
 }
 
+// packages/mcps/src/mcp/tools/e2e/test-state-upstream.ts
+function buildTestChainingCatalogPath(projectId) {
+  return `${MUGGLE_TEST_PREFIX}/projects/${projectId}/test-chaining/catalog`;
+}
+function buildTestStatesPath(projectId) {
+  return `${MUGGLE_TEST_PREFIX}/projects/${projectId}/test-chaining/states`;
+}
+function buildTestRolesPath(projectId) {
+  return `${MUGGLE_TEST_PREFIX}/projects/${projectId}/test-chaining/roles`;
+}
+function buildTestChainingSettingsPath(projectId) {
+  return `${MUGGLE_TEST_PREFIX}/projects/${projectId}/test-chaining/settings`;
+}
+function buildTestChainingPlanPath(projectId) {
+  return `${MUGGLE_TEST_PREFIX}/projects/${projectId}/test-chaining/plan`;
+}
+function pickTestCaseTestStateFields(testStateFields) {
+  const requestFields = {};
+  if (testStateFields.actsAsRoleIds !== void 0) requestFields.actsAsRoleIds = testStateFields.actsAsRoleIds;
+  if (testStateFields.prerequisiteStateId !== void 0) {
+    requestFields.prerequisiteStateId = testStateFields.prerequisiteStateId;
+  }
+  if (testStateFields.resultStateId !== void 0) requestFields.resultStateId = testStateFields.resultStateId;
+  return requestFields;
+}
+
 // packages/mcps/src/mcp/tools/e2e/tool-registry.ts
 var getWorkflowTimeoutMs = () => getConfig().e2e.workflowTimeoutMs;
 var projectTools = [
@@ -8000,7 +8110,7 @@ var testCaseTools = [
   },
   {
     name: "muggle-remote-test-case-create",
-    description: "Create a new test case for a use case.",
+    description: "Create a new test case for a use case. Optionally place it on the project's test states: the roles it acts as (actsAsRoleIds), the state it needs to start (prerequisiteStateId) and the state it leaves when it passes (resultStateId), using ids from muggle-remote-test-chaining-catalog-get.",
     inputSchema: TestCaseCreateInputSchema,
     dashboardLink: linkToInputProjectPage("test-cases" /* TestCases */),
     mapToUpstream: (input) => {
@@ -8025,18 +8135,19 @@ var testCaseTools = [
           ...toolInput.pinnedAccount !== void 0 && { pinnedAccount: toolInput.pinnedAccount },
           ...toolInput.alwaysStartsFreshSession !== void 0 && {
             alwaysStartsFreshSession: toolInput.alwaysStartsFreshSession
-          }
+          },
+          ...pickTestCaseTestStateFields(toolInput)
         }
       };
     }
   },
   {
     name: "muggle-remote-test-case-update",
-    description: "Directly update an existing test case's fields. Pass only the fields you want to change \u2014 others are left untouched. Use this for cheap, deterministic edits (rename, change status/priority, fix expected result, add tags, set or clear the pinned account with pinnedAccount, set whether every run starts from a fresh session with alwaysStartsFreshSession; pinned account and fresh session apply from the next script generation, not to existing scripts). Does not touch the associated test script \u2014 script regeneration is a separate workflow (muggle-remote-workflow-start-test-script-generation).",
+    description: "Directly update an existing test case's fields. Pass only the fields you want to change \u2014 others are left untouched. Use this for cheap, deterministic edits (rename, change status/priority, fix expected result, add tags, set or clear the pinned account with pinnedAccount, set whether every run starts from a fresh session with alwaysStartsFreshSession; pinned account and fresh session apply from the next script generation, not to existing scripts; set the roles it acts as and its prerequisite and result test states with actsAsRoleIds, prerequisiteStateId and resultStateId, null clearing each). Does not touch the associated test script \u2014 script regeneration is a separate workflow (muggle-remote-workflow-start-test-script-generation).",
     inputSchema: TestCaseUpdateInputSchema,
     mapToUpstream: (input) => {
       const toolInput = input;
-      const body = { id: toolInput.testCaseId };
+      const body = { id: toolInput.testCaseId, ...pickTestCaseTestStateFields(toolInput) };
       if (toolInput.title !== void 0) body.title = toolInput.title;
       if (toolInput.description !== void 0) body.description = toolInput.description;
       if (toolInput.goal !== void 0) body.goal = toolInput.goal;
@@ -8085,6 +8196,146 @@ var testCaseTools = [
     }
   }
 ];
+var testStateTools = [
+  {
+    name: "muggle-remote-test-chaining-catalog-get",
+    description: "Get a project's test states and roles. A state is a named set of facts about one account set's data that tests start from and leave, each fact scoped to the account set or account-wide, with a quick read-only check. A role is a kind of account a test acts as, CREATED by Muggle or SUPPLIED by the customer. Also returns freshAccountSetStateId (the state a newly formed account set starts in) and parallelAccountSetLimit; a project with none saved returns an empty catalog. Response shape: { projectId, states: [{ id, name, facts: [{ statement, scope, roleId }], check }], roles: [{ id, name, source, allowsChanges }], freshAccountSetStateId, parallelAccountSetLimit, updatedAt }.",
+    inputSchema: TestChainingCatalogGetInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      return {
+        method: "GET",
+        path: buildTestChainingCatalogPath(toolInput.projectId)
+      };
+    }
+  },
+  {
+    name: "muggle-remote-test-chaining-settings-update",
+    description: "Set a project's freshAccountSetStateId (the state a newly formed account set starts in, such as 'New patient'; null clears it) and parallelAccountSetLimit (the most account sets a batch runs at once). Omitted fields stay unchanged. Returns the whole catalog, as muggle-remote-test-chaining-catalog-get does.",
+    inputSchema: TestChainingSettingsUpdateInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      const body = {};
+      if (toolInput.freshAccountSetStateId !== void 0) body.freshAccountSetStateId = toolInput.freshAccountSetStateId;
+      if (toolInput.parallelAccountSetLimit !== void 0) body.parallelAccountSetLimit = toolInput.parallelAccountSetLimit;
+      return {
+        method: "PUT",
+        path: buildTestChainingSettingsPath(toolInput.projectId),
+        body
+      };
+    }
+  },
+  {
+    name: "muggle-remote-test-state-create",
+    description: "Create a test state: a named set of facts about one account set's data, such as 'Booked' or 'Message unread'. List every fact the state's tests rely on, each scoped to the account set (only this account set's data, e.g. the patient's thread with the doctor) or account-wide (one account as a whole, e.g. the doctor's unread count, naming that account's role). check is a quick read-only check that an account set is in the state. Reuse a matching state from muggle-remote-test-chaining-catalog-get before creating one. Returns the created state.",
+    inputSchema: TestStateCreateInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      const body = { name: toolInput.name };
+      if (toolInput.facts !== void 0) body.facts = toolInput.facts;
+      if (toolInput.check !== void 0) body.check = toolInput.check;
+      return {
+        method: "POST",
+        path: buildTestStatesPath(toolInput.projectId),
+        body
+      };
+    }
+  },
+  {
+    name: "muggle-remote-test-state-update",
+    description: "Update a test state's name, facts or check. Omitted fields stay unchanged; facts replaces the state's whole fact list. Returns the updated state.",
+    inputSchema: TestStateUpdateInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      const body = {};
+      if (toolInput.name !== void 0) body.name = toolInput.name;
+      if (toolInput.facts !== void 0) body.facts = toolInput.facts;
+      if (toolInput.check !== void 0) body.check = toolInput.check;
+      return {
+        method: "PUT",
+        path: `${buildTestStatesPath(toolInput.projectId)}/${toolInput.stateId}`,
+        body
+      };
+    }
+  },
+  {
+    name: "muggle-remote-test-state-delete",
+    description: "Delete a test state. Rejected while a test case uses it as its prerequisite or result state, or while it is the project's freshAccountSetStateId.",
+    inputSchema: TestStateDeleteInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      return {
+        method: "DELETE",
+        path: `${buildTestStatesPath(toolInput.projectId)}/${toolInput.stateId}`
+      };
+    },
+    mapFromUpstream: (_response, input) => {
+      const toolInput = input;
+      return { stateId: toolInput.stateId, isDeleted: true };
+    }
+  },
+  {
+    name: "muggle-remote-test-role-create",
+    description: "Create a test role: a kind of account a test acts as, such as Patient or Doctor; two accounts of one kind in one test are two roles. CREATED: Muggle signs each account set's account up itself, and it is disposable. SUPPLIED: the customer provides the account, which is never used up; allowsChanges lets tests change its account-wide facts when a proven way back restores them. Returns the created role.",
+    inputSchema: TestRoleCreateInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      const body = { name: toolInput.name, source: toolInput.source };
+      if (toolInput.allowsChanges !== void 0) body.allowsChanges = toolInput.allowsChanges;
+      return {
+        method: "POST",
+        path: buildTestRolesPath(toolInput.projectId),
+        body
+      };
+    }
+  },
+  {
+    name: "muggle-remote-test-role-update",
+    description: "Update a test role's name, source or allowsChanges. Omitted fields stay unchanged. Returns the updated role.",
+    inputSchema: TestRoleUpdateInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      const body = {};
+      if (toolInput.name !== void 0) body.name = toolInput.name;
+      if (toolInput.source !== void 0) body.source = toolInput.source;
+      if (toolInput.allowsChanges !== void 0) body.allowsChanges = toolInput.allowsChanges;
+      return {
+        method: "PUT",
+        path: `${buildTestRolesPath(toolInput.projectId)}/${toolInput.roleId}`,
+        body
+      };
+    }
+  },
+  {
+    name: "muggle-remote-test-role-delete",
+    description: "Delete a test role. Rejected while a test case acts as it or a state's account-wide fact names it.",
+    inputSchema: TestRoleDeleteInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      return {
+        method: "DELETE",
+        path: `${buildTestRolesPath(toolInput.projectId)}/${toolInput.roleId}`
+      };
+    },
+    mapFromUpstream: (_response, input) => {
+      const toolInput = input;
+      return { roleId: toolInput.roleId, isDeleted: true };
+    }
+  },
+  {
+    name: "muggle-remote-test-chaining-plan-get",
+    description: "Preview how the project's test cases would run on account sets. Read-only: it plans from the saved states, roles and test cases and does not run anything. Returns the floor (fewest account sets that can run every placed test), each account set's route in order with setup runs (replays that only reach a state, with no verdict) and the one-way test that forced each account set beyond the first, the groups of states, Not allowed tests that would change a supplied account, tests whose prerequisite state no test reaches, and test cases with no states yet (they run as today). Response shape: { floor, batchLength, totalRuns, setupRuns, account sets: [{ accountSetNumber, slotNumber, steps: [{ testCaseId, prerequisiteStateId, resultStateId, isSetup, heldSuppliedRoleIds }], endStateId, isUsedUp, forcedByTestCaseId }], groups: [{ groupNumber, stateIds, isTerminal }], notAllowed: [{ testCaseId, reason, suppliedRoleIds }], missingState: [{ testCaseId, prerequisiteStateId }], unplacedTestCaseIds, unrestoredSuppliedRoleIds }.",
+    inputSchema: TestChainingPlanGetInputSchema,
+    mapToUpstream: (input) => {
+      const toolInput = input;
+      return {
+        method: "GET",
+        path: buildTestChainingPlanPath(toolInput.projectId),
+        queryParams: { isProduction: toolInput.isProduction }
+      };
+    }
+  }
+];
 var bulkPreviewTools = [
   {
     name: "muggle-remote-bulk-preview-job-get",
@@ -9368,6 +9619,7 @@ var allQaToolDefinitions = [
   ...projectTools,
   ...useCaseTools,
   ...testCaseTools,
+  ...testStateTools,
   ...bulkPreviewTools,
   ...testScriptTools,
   ...actionScriptTools,
@@ -10708,11 +10960,24 @@ __export(e2e_exports2, {
   TestCaseGetInputSchema: () => TestCaseGetInputSchema,
   TestCaseListByUseCaseInputSchema: () => TestCaseListByUseCaseInputSchema,
   TestCaseListInputSchema: () => TestCaseListInputSchema,
+  TestCaseTestStateFieldsSchema: () => TestCaseTestStateFieldsSchema,
   TestCaseUpdateInputSchema: () => TestCaseUpdateInputSchema,
+  TestChainingCatalogGetInputSchema: () => TestChainingCatalogGetInputSchema,
+  TestChainingPlanGetInputSchema: () => TestChainingPlanGetInputSchema,
+  TestChainingSettingsUpdateInputSchema: () => TestChainingSettingsUpdateInputSchema,
   TestPlanGraphRebuildInputSchema: () => TestPlanGraphRebuildInputSchema,
+  TestRoleCreateInputSchema: () => TestRoleCreateInputSchema,
+  TestRoleDeleteInputSchema: () => TestRoleDeleteInputSchema,
+  TestRoleSourceEnum: () => TestRoleSourceEnum,
+  TestRoleUpdateInputSchema: () => TestRoleUpdateInputSchema,
   TestScriptDeleteInputSchema: () => TestScriptDeleteInputSchema,
   TestScriptGetInputSchema: () => TestScriptGetInputSchema,
   TestScriptListInputSchema: () => TestScriptListInputSchema,
+  TestStateCreateInputSchema: () => TestStateCreateInputSchema,
+  TestStateDeleteInputSchema: () => TestStateDeleteInputSchema,
+  TestStateFactSchema: () => TestStateFactSchema,
+  TestStateFactScopeEnum: () => TestStateFactScopeEnum,
+  TestStateUpdateInputSchema: () => TestStateUpdateInputSchema,
   TokenPackageIdSchema: () => TokenPackageIdSchema,
   TokenUsageFilterTypeSchema: () => TokenUsageFilterTypeSchema,
   UseCaseCandidatesApproveInputSchema: () => UseCaseCandidatesApproveInputSchema,

package/dist/cli.js

@@ -1,6 +1,6 @@
 #!/usr/bin/env node
-import { runCli } from './chunk-J2DEXYH7.js';
-import './chunk-YLQSNSSM.js';
+import { runCli } from './chunk-W3HJBL5U.js';
+import './chunk-XUANXZ2O.js';
 
 // src/cli/main.ts
 runCli().catch((error) => {

package/dist/index.js

@@ -1,2 +1,2 @@
-export { src_exports as commands, createUnifiedMcpServer, server_exports as server } from './chunk-J2DEXYH7.js';
-export { createChildLogger, e2e_exports as e2e, getConfig, getLocalQaTools, getLogger, getQaTools, local_exports as localQa, mcp_exports as mcp, e2e_exports as qa, src_exports as shared } from './chunk-YLQSNSSM.js';
+export { src_exports as commands, createUnifiedMcpServer, server_exports as server } from './chunk-W3HJBL5U.js';
+export { createChildLogger, e2e_exports as e2e, getConfig, getLocalQaTools, getLogger, getQaTools, local_exports as localQa, mcp_exports as mcp, e2e_exports as qa, src_exports as shared } from './chunk-XUANXZ2O.js';

package/dist/plugin/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "muggle",
   "description": "Real-browser E2E acceptance for your coding agent: test the flows you changed and open the PR with screenshots. Built for flows other tools can't reach: managed test identities with live inboxes (magic links, email OTP and 2FA, verification mail, password resets), an encrypted secret store for login and test credentials, and a test graph that runs prerequisite flows first. Plain-English tests become replayable scripts on localhost or preview URLs.",
-  "version": "5.31.0-staging.268",
+  "version": "5.31.0-staging.270",
   "author": {
     "name": "Muggle AI",
     "email": "support@muggle-ai.com"

package/dist/plugin/.codex-mcp.json

@@ -4,7 +4,7 @@
       "command": "npx",
       "args": [
         "-y",
-        "@muggleai/works@5.31.0-staging.268",
+        "@muggleai/works@5.31.0-staging.270",
         "serve"
       ],
       "env": {

package/dist/plugin/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "muggle",
-  "version": "5.31.0-staging.268",
+  "version": "5.31.0-staging.270",
   "description": "Real-browser E2E acceptance for your coding agent: test the flows you changed and open the PR with screenshots. Built for flows other tools can't reach: managed test identities with live inboxes (magic links, email OTP and 2FA, verification mail, password resets), an encrypted secret store for login and test credentials, and a test graph that runs prerequisite flows first. Plain-English tests become replayable scripts on localhost or preview URLs.",
   "author": {
     "name": "Muggle AI",

package/dist/plugin/.cursor-plugin/plugin.json

@@ -2,7 +2,7 @@
   "name": "muggle",
   "displayName": "Muggle AI",
   "description": "Real-browser E2E acceptance for your coding agent: test the flows you changed and open the PR with screenshots. Built for flows other tools can't reach: managed test identities with live inboxes (magic links, email OTP and 2FA, verification mail, password resets), an encrypted secret store for login and test credentials, and a test graph that runs prerequisite flows first. Plain-English tests become replayable scripts on localhost or preview URLs.",
-  "version": "5.31.0-staging.268",
+  "version": "5.31.0-staging.270",
   "author": {
     "name": "Muggle AI",
     "email": "support@muggle-ai.com"

package/dist/plugin/.mcp.json

@@ -4,7 +4,7 @@
       "command": "npx",
       "args": [
         "-y",
-        "@muggleai/works@5.31.0-staging.268",
+        "@muggleai/works@5.31.0-staging.270",
         "serve"
       ]
     }

package/dist/plugin/skills/do/fix-ci.md

@@ -33,6 +33,8 @@ Then increment `last_seen.ci_fix_attempts[red_sha]` — before any fix, re-run o
 
 The increment opens a cycle this turn carries through Step 5 or Step 6 — it never stands in for one. The single exception is a red `muggle-walkthrough` whose E2E run for **this** PR is already queued or running (its `state.md` shows `E2E dispatch: queued` or `running`): that run is the cycle, and its report settles the comment. Writing the count for a PR with no such run, or because another PR's run is in flight, clears the turn-end gate while the PR stays red; the watch never asks again until the head moves, and the PR can merge red.
 
+If the increment takes the count past 3, this head's budget was spent before the cycle began: make no fix, no re-run and no push — go straight to Step 6. A fixable failure does not reopen a spent budget.
+
 ### Step 2 — Map each failing check to a local command
 
 Per [`../_shared/ci-check-to-command.md`](../_shared/ci-check-to-command.md). Fix the in-scope checks in the working tree; record out-of-scope checks for escalation (Step 6). A red `muggle-walkthrough` check takes Step 2.5; a red non-deterministic check takes Step 2.6.
@@ -84,7 +86,7 @@ Respawn the watcher per [`respawn-watcher.md`](respawn-watcher.md). CI on the ne
 
 ### Step 6 — Escalate (budget spent or out of scope)
 
-When the failing checks are all out of scope, a non-deterministic check is red again after its re-run on this head (Step 2.6), or this is the head's third cycle (`ci_fix_attempts[red_sha]` reached 3 in Step 1) and it could not push a fix:
+When the failing checks are all out of scope, a non-deterministic check is red again after its re-run on this head (Step 2.6), this is the head's third cycle (`ci_fix_attempts[red_sha]` reached 3 in Step 1) and it could not push a fix, or Step 1 found the budget already spent:
 
 1. Add `red_sha` to `last_seen.ci_escalated_shas` so the SHA is not re-fixed. This cycle's Step 1 attempt is what makes the entry count.
 2. Emit one terminal message naming the unresolved checks.

package/dist/release-manifest.json

@@ -1,7 +1,7 @@
 {
   "release": "5.26.0",
-  "buildId": "run-268-1",
-  "commitSha": "034abfdee292104743ae296d7c24aa85de74066e",
-  "buildTime": "2026-10-31T20:23:14Z",
+  "buildId": "run-270-1",
+  "commitSha": "09e34fedf072b7a5ab4fdce746bd0c0bfbe53c4a",
+  "buildTime": "2026-10-31T23:24:31Z",
   "serviceName": "muggle-ai-works-mcp"
 }

package/dist/{src-7OY7H63R.js → src-J6AIYTLA.js}

@@ -1 +1 @@
-export { ACTION_SCRIPT_FILE_NAME, BASE_REMOTE_NAME, BLOCKING_FRESHNESS_VERDICTS, BRANCH_CREATION_REFLOG_PREFIX, BUSY_EXIT_CODE, CLAUDE_CODE_AUTH_STATUS_ARGS, CLAUDE_CODE_CHECK_TIMEOUT_MS, CLAUDE_CODE_EXECUTABLE, CLAUDE_CODE_EXECUTABLE_NOT_FOUND_CODE, CLAUDE_CODE_SESSION_LINK_ENV_VARS, CLAUDE_CODE_VERSION_ARGS, CheckoutFreshnessVerdict, ClaudeCodeReadiness, DEFAULT_LOCAL_LLM_BASE_URL, DEFAULT_PREFERENCES, ELECTRON_APP_EXECUTABLE_PATH_SEGMENTS, ElectronAppPlatform, ElectronAppReleaseStream, HTML_VIEW_OFFER_LINE, LLM_CREDENTIALS_FILE_MODE, LLM_CREDENTIALS_FILE_NAME, LLM_SOURCE_HEADER, LLM_SOURCE_OPTIONS, LLM_SOURCE_OWNED_ENV_NAMES, LLM_SOURCE_PROMPT, LLM_SOURCE_SCOPE_NOTE, LlmSource, MASKED_API_KEY_PREFIX, MASKED_API_KEY_VISIBLE_CHARS, MAX_LISTED_INHERITED_COMMITS, ONBOARDING_BLANKET_OPTIONS, ONBOARDING_CHOICE_GROUPS, ONBOARDING_MAX_OFFERS, ONBOARDING_PRIMER_BULLETS, ONBOARDING_PRIMER_HEADLINE, ONBOARDING_TOGGLE_GROUPS, ONBOARDING_TOGGLE_PROMPT, ONBOARDING_UNSELECTED_TOGGLE_VALUE, OnboardingBlanketChoice, OnboardingGroupKind, PREFERENCES_FILE_NAME, PREFERENCES_PROJECT_DIR_NAME, PREFERENCES_SCHEMA, PREFERENCES_VERSION, PREFERENCE_ALLOWED_VALUES, PreferenceKey, PreferenceValue, ProjectPreferencesReconcileOutcome, RUN_SESSIONS_DIR_NAME, RUN_VIEW_DASHBOARD_LINK_LABEL, RunViewError, RuntimeTarget, SCREENSHOT_DIR_RELATIVE_PATH, SCREENSHOT_FILE_EXTENSION, StackCheckVerdict, StackClaimVerdict, StackLeaseEndReason, StackMismatchKind, StackPortAction, WATCHER_LIFETIME_SECONDS, WATCHER_LIFETIME_UNBOUNDED_SECONDS, applyLlmSourceSelection, applyOnboardingAnswers, assertDeviceCodeClientProvisioned, buildElectronAppChecksumsUrl, buildElectronAppDirName, buildElectronAppReleaseAssetUrl, buildElectronAppReleaseTag, buildLlmSourcePlan, buildOnboardingPlan, buildToolInputJsonSchema, calculateFileChecksum, checkServiceStack, claimServiceStack, compareQueuePlaces, completeOnboarding, createApiKeyWithToken, createChildLogger, decideStackClaim, decideStackReap, decideStackRelease, decideStackRunEnd, decideStackRunStart, deleteApiKeyData, deleteCredentials, describeLlmSourceStatus, describeStackMismatch, e2e_exports as e2e, evaluateCheckoutFreshness, findVacatedNeeds, finishStackRun, formatPreferencesOneLiner, getActiveElectronAppReleaseStream, getActiveRuntimeTarget, getApiKey, getApiKeyFilePath, getAuthService, getBundledElectronAppVersion, getCallerCredentials, getCallerCredentialsAsync, getChecksumForPlatform, getConfig, getCredentialsFilePath, getDashboardBaseUrl, getDataDir2 as getDataDir, getDownloadBaseUrl, getElectronAppChecksums, getElectronAppDir, getElectronAppReleaseTagPrefix, getElectronAppSignedFromVersion, getElectronAppVersion, getElectronAppVersionSource, getLlmSourceOption, getLocalQaTools, getLogger, getOnboardingToggleKeys, getPlatformKey, getQaTools, getReleaseSignerIdentityUri, getRunSessionsDir, getValidApiKeyData, getValidCredentials, hasApiKey, isElectronAppInstalled, isElectronAppPlatform, isFirstRun, isValidLocalBaseUrl, loadApiKeyData, loadCredentials, local_exports as localQa, lookUpCheckoutPullRequest, markOnboardingCompleted, maskLlmApiKey, mcp_exports as mcp, needsConflict, needsOnboarding, openBrowserUrl, parseElectronAppDirName, performLogin, performLogout, pollDeviceCode, portOfUrl, e2e_exports as qa, readPreferencesMetadata, readRunView, readServiceStackStatus, reapServiceStack, reconcileProjectPreferences, recordOnboardingSkip, releaseServiceStack, renderRunViewHtml, renderRunViewText, resetConfig, resetLogger, resetPreference, resolveActiveProfile, resolveActiveReleaseStream, resolveActiveReleaseTagPrefix, resolveBlanketPreferences, resolveElectronAppExecutablePath, resolveElectronAppPathOrNull, resolveLlmSourceStatus, resolveOnboardingAnswers, resolvePreferences, resolveRunSessionPath, resolveRuntimeTarget, resolveTogglePreferences, runGitCommand, saveApiKey, saveApiKeyData, saveCredentials, startDeviceCodeFlow, startStackRun, stopVacatedServices, toDisplayRef, toolRequiresAuth, validatePreference, verifyFileChecksum, writeLlmEnvBlock, writePreferences, writePreferencesWithMetadata, writeRunViewHtml } from './chunk-YLQSNSSM.js';
+export { ACTION_SCRIPT_FILE_NAME, BASE_REMOTE_NAME, BLOCKING_FRESHNESS_VERDICTS, BRANCH_CREATION_REFLOG_PREFIX, BUSY_EXIT_CODE, CLAUDE_CODE_AUTH_STATUS_ARGS, CLAUDE_CODE_CHECK_TIMEOUT_MS, CLAUDE_CODE_EXECUTABLE, CLAUDE_CODE_EXECUTABLE_NOT_FOUND_CODE, CLAUDE_CODE_SESSION_LINK_ENV_VARS, CLAUDE_CODE_VERSION_ARGS, CheckoutFreshnessVerdict, ClaudeCodeReadiness, DEFAULT_LOCAL_LLM_BASE_URL, DEFAULT_PREFERENCES, ELECTRON_APP_EXECUTABLE_PATH_SEGMENTS, ElectronAppPlatform, ElectronAppReleaseStream, HTML_VIEW_OFFER_LINE, LLM_CREDENTIALS_FILE_MODE, LLM_CREDENTIALS_FILE_NAME, LLM_SOURCE_HEADER, LLM_SOURCE_OPTIONS, LLM_SOURCE_OWNED_ENV_NAMES, LLM_SOURCE_PROMPT, LLM_SOURCE_SCOPE_NOTE, LlmSource, MASKED_API_KEY_PREFIX, MASKED_API_KEY_VISIBLE_CHARS, MAX_LISTED_INHERITED_COMMITS, ONBOARDING_BLANKET_OPTIONS, ONBOARDING_CHOICE_GROUPS, ONBOARDING_MAX_OFFERS, ONBOARDING_PRIMER_BULLETS, ONBOARDING_PRIMER_HEADLINE, ONBOARDING_TOGGLE_GROUPS, ONBOARDING_TOGGLE_PROMPT, ONBOARDING_UNSELECTED_TOGGLE_VALUE, OnboardingBlanketChoice, OnboardingGroupKind, PREFERENCES_FILE_NAME, PREFERENCES_PROJECT_DIR_NAME, PREFERENCES_SCHEMA, PREFERENCES_VERSION, PREFERENCE_ALLOWED_VALUES, PreferenceKey, PreferenceValue, ProjectPreferencesReconcileOutcome, RUN_SESSIONS_DIR_NAME, RUN_VIEW_DASHBOARD_LINK_LABEL, RunViewError, RuntimeTarget, SCREENSHOT_DIR_RELATIVE_PATH, SCREENSHOT_FILE_EXTENSION, StackCheckVerdict, StackClaimVerdict, StackLeaseEndReason, StackMismatchKind, StackPortAction, WATCHER_LIFETIME_SECONDS, WATCHER_LIFETIME_UNBOUNDED_SECONDS, applyLlmSourceSelection, applyOnboardingAnswers, assertDeviceCodeClientProvisioned, buildElectronAppChecksumsUrl, buildElectronAppDirName, buildElectronAppReleaseAssetUrl, buildElectronAppReleaseTag, buildLlmSourcePlan, buildOnboardingPlan, buildToolInputJsonSchema, calculateFileChecksum, checkServiceStack, claimServiceStack, compareQueuePlaces, completeOnboarding, createApiKeyWithToken, createChildLogger, decideStackClaim, decideStackReap, decideStackRelease, decideStackRunEnd, decideStackRunStart, deleteApiKeyData, deleteCredentials, describeLlmSourceStatus, describeStackMismatch, e2e_exports as e2e, evaluateCheckoutFreshness, findVacatedNeeds, finishStackRun, formatPreferencesOneLiner, getActiveElectronAppReleaseStream, getActiveRuntimeTarget, getApiKey, getApiKeyFilePath, getAuthService, getBundledElectronAppVersion, getCallerCredentials, getCallerCredentialsAsync, getChecksumForPlatform, getConfig, getCredentialsFilePath, getDashboardBaseUrl, getDataDir2 as getDataDir, getDownloadBaseUrl, getElectronAppChecksums, getElectronAppDir, getElectronAppReleaseTagPrefix, getElectronAppSignedFromVersion, getElectronAppVersion, getElectronAppVersionSource, getLlmSourceOption, getLocalQaTools, getLogger, getOnboardingToggleKeys, getPlatformKey, getQaTools, getReleaseSignerIdentityUri, getRunSessionsDir, getValidApiKeyData, getValidCredentials, hasApiKey, isElectronAppInstalled, isElectronAppPlatform, isFirstRun, isValidLocalBaseUrl, loadApiKeyData, loadCredentials, local_exports as localQa, lookUpCheckoutPullRequest, markOnboardingCompleted, maskLlmApiKey, mcp_exports as mcp, needsConflict, needsOnboarding, openBrowserUrl, parseElectronAppDirName, performLogin, performLogout, pollDeviceCode, portOfUrl, e2e_exports as qa, readPreferencesMetadata, readRunView, readServiceStackStatus, reapServiceStack, reconcileProjectPreferences, recordOnboardingSkip, releaseServiceStack, renderRunViewHtml, renderRunViewText, resetConfig, resetLogger, resetPreference, resolveActiveProfile, resolveActiveReleaseStream, resolveActiveReleaseTagPrefix, resolveBlanketPreferences, resolveElectronAppExecutablePath, resolveElectronAppPathOrNull, resolveLlmSourceStatus, resolveOnboardingAnswers, resolvePreferences, resolveRunSessionPath, resolveRuntimeTarget, resolveTogglePreferences, runGitCommand, saveApiKey, saveApiKeyData, saveCredentials, startDeviceCodeFlow, startStackRun, stopVacatedServices, toDisplayRef, toolRequiresAuth, validatePreference, verifyFileChecksum, writeLlmEnvBlock, writePreferences, writePreferencesWithMetadata, writeRunViewHtml } from './chunk-XUANXZ2O.js';

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@muggleai/works",
   "mcpName": "io.github.multiplex-ai/muggle",
-  "version": "5.31.0-staging.268",
+  "version": "5.31.0-staging.270",
   "description": "13 years of Microsoft hyperscale engineering, packaged as a harness for your AI coding agent.",
   "type": "module",
   "main": "dist/index.js",

package/plugin/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "muggle",
   "description": "Real-browser E2E acceptance for your coding agent: test the flows you changed and open the PR with screenshots. Built for flows other tools can't reach: managed test identities with live inboxes (magic links, email OTP and 2FA, verification mail, password resets), an encrypted secret store for login and test credentials, and a test graph that runs prerequisite flows first. Plain-English tests become replayable scripts on localhost or preview URLs.",
-  "version": "5.31.0-staging.268",
+  "version": "5.31.0-staging.270",
   "author": {
     "name": "Muggle AI",
     "email": "support@muggle-ai.com"

package/plugin/.codex-mcp.json

@@ -4,7 +4,7 @@
       "command": "npx",
       "args": [
         "-y",
-        "@muggleai/works@5.31.0-staging.268",
+        "@muggleai/works@5.31.0-staging.270",
         "serve"
       ],
       "env": {

package/plugin/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "muggle",
-  "version": "5.31.0-staging.268",
+  "version": "5.31.0-staging.270",
   "description": "Real-browser E2E acceptance for your coding agent: test the flows you changed and open the PR with screenshots. Built for flows other tools can't reach: managed test identities with live inboxes (magic links, email OTP and 2FA, verification mail, password resets), an encrypted secret store for login and test credentials, and a test graph that runs prerequisite flows first. Plain-English tests become replayable scripts on localhost or preview URLs.",
   "author": {
     "name": "Muggle AI",

package/plugin/.cursor-plugin/plugin.json

@@ -2,7 +2,7 @@
   "name": "muggle",
   "displayName": "Muggle AI",
   "description": "Real-browser E2E acceptance for your coding agent: test the flows you changed and open the PR with screenshots. Built for flows other tools can't reach: managed test identities with live inboxes (magic links, email OTP and 2FA, verification mail, password resets), an encrypted secret store for login and test credentials, and a test graph that runs prerequisite flows first. Plain-English tests become replayable scripts on localhost or preview URLs.",
-  "version": "5.31.0-staging.268",
+  "version": "5.31.0-staging.270",
   "author": {
     "name": "Muggle AI",
     "email": "support@muggle-ai.com"

package/plugin/.mcp.json

@@ -4,7 +4,7 @@
       "command": "npx",
       "args": [
         "-y",
-        "@muggleai/works@5.31.0-staging.268",
+        "@muggleai/works@5.31.0-staging.270",
         "serve"
       ]
     }

package/plugin/skills/do/fix-ci.md

@@ -33,6 +33,8 @@ Then increment `last_seen.ci_fix_attempts[red_sha]` — before any fix, re-run o
 
 The increment opens a cycle this turn carries through Step 5 or Step 6 — it never stands in for one. The single exception is a red `muggle-walkthrough` whose E2E run for **this** PR is already queued or running (its `state.md` shows `E2E dispatch: queued` or `running`): that run is the cycle, and its report settles the comment. Writing the count for a PR with no such run, or because another PR's run is in flight, clears the turn-end gate while the PR stays red; the watch never asks again until the head moves, and the PR can merge red.
 
+If the increment takes the count past 3, this head's budget was spent before the cycle began: make no fix, no re-run and no push — go straight to Step 6. A fixable failure does not reopen a spent budget.
+
 ### Step 2 — Map each failing check to a local command
 
 Per [`../_shared/ci-check-to-command.md`](../_shared/ci-check-to-command.md). Fix the in-scope checks in the working tree; record out-of-scope checks for escalation (Step 6). A red `muggle-walkthrough` check takes Step 2.5; a red non-deterministic check takes Step 2.6.
@@ -84,7 +86,7 @@ Respawn the watcher per [`respawn-watcher.md`](respawn-watcher.md). CI on the ne
 
 ### Step 6 — Escalate (budget spent or out of scope)
 
-When the failing checks are all out of scope, a non-deterministic check is red again after its re-run on this head (Step 2.6), or this is the head's third cycle (`ci_fix_attempts[red_sha]` reached 3 in Step 1) and it could not push a fix:
+When the failing checks are all out of scope, a non-deterministic check is red again after its re-run on this head (Step 2.6), this is the head's third cycle (`ci_fix_attempts[red_sha]` reached 3 in Step 1) and it could not push a fix, or Step 1 found the budget already spent:
 
 1. Add `red_sha` to `last_seen.ci_escalated_shas` so the SHA is not re-fixed. This cycle's Step 1 attempt is what makes the entry count.
 2. Emit one terminal message naming the unresolved checks.