vibe-coding-master 0.7.20 → 0.7.22

package/README.md

@@ -9,7 +9,7 @@ VCM helps you keep one task organized across dedicated roles:
 - Architect
 - Coder
 - Tester
-- optional Gate Reviewer
+- optional Reviewer
 
 It runs locally, connects to a local Git repository, creates a task branch and
 worktree, starts embedded Claude Code terminals, and manages handoffs,
@@ -133,7 +133,7 @@ If you want VCM app state to survive container rebuilds, set:
 7. Review the harness/bootstrap commit diff.
 8. Start the role sessions, or use the saved launch template / one-click start.
 9. Talk mostly to Project Manager.
-10. Let PM route work to Architect, Coder, Tester, and Gate Reviewer when
+10. Let PM route work to Architect, Coder, Tester, and Reviewer when
     enabled.
 11. Review the final result and close the task when finished.
 
@@ -290,7 +290,7 @@ The global launch template stores per-role defaults for the five workflow roles:
 - auto orchestration
 
 One-click start launches the four core roles. If any Gate Review Gate is enabled,
-it also launches Gate Reviewer.
+it also launches Reviewer.
 
 Translator and Harness Engineer are tool roles. They are controlled from their
 own panels, not from the main role tab bar or the one-click launch template.
@@ -331,14 +331,14 @@ Available gates:
 - Validation adequacy
 - Code diff
 
-When a gate is enabled, VCM uses Gate Reviewer as an independent review role.
-Gate Reviewer reviews artifacts and diffs, writes a gate report, and returns
+When a gate is enabled, VCM uses Reviewer as an independent review role.
+Reviewer reviews artifacts and diffs, writes a gate report, and returns
 only:
 
 - `approve`
 - `request_changes`
 
-Gate Reviewer does not run tests and does not choose owners or fixes. PM routes
+Reviewer does not run tests and does not choose owners or fixes. PM routes
 findings back to the responsible role.
 
 ## Translation
@@ -405,7 +405,7 @@ Gateway can:
 - translate mobile messages when Gateway translation is enabled
 
 Gateway does not expose the embedded terminal and does not send directly to
-Architect, Coder, Tester, or Gate Reviewer.
+Architect, Coder, Tester, or Reviewer.
 
 Common commands:
 
@@ -462,7 +462,7 @@ files.
 
 `Auto memory` is the switch for the entire automated memory workflow. During
 Review Task Harness after Final Acceptance, Project Manager, Architect, Coder,
-Tester, and an enabled Gate Reviewer submit proposals in sequence through
+Tester, and an enabled Reviewer submit proposals in sequence through
 `vcm-propose-memory`. Harness Engineer verifies and consolidates them before VCM
 applies the result. Roles cannot edit active memory directly.
 
@@ -484,7 +484,7 @@ Final Acceptance
 ```
 
 Memory proposal prompts sent to Project Manager, Architect, Coder, Tester, and
-an enabled Gate Reviewer use their normal task sessions and participate in
+an enabled Reviewer use their normal task sessions and participate in
 Round/Turn tracking. Harness Engineer review and retrospective work remain tool
 role activity and do not participate in Round completion.
 

package/dist/backend/cli/install-vcm-harness.js

@@ -8,7 +8,7 @@ import { renderArchitectHarnessRules } from "../templates/harness/architect-agen
 import { renderCoderHarnessRules } from "../templates/harness/coder-agent.js";
 import { renderCoderWorkerHarnessRules } from "../templates/harness/coder-worker-agent.js";
 import { renderArchitectScaffoldWorkerHarnessRules } from "../templates/harness/architect-scaffold-worker-agent.js";
-import { renderGateReviewerAgentRules, renderRequestGateReviewTool, renderTranslatorAgentRules, renderVcmGateReviewSkillRules } from "../templates/harness/gate-review.js";
+import { renderReviewerAgentRules, renderRequestGateReviewTool, renderTranslatorAgentRules, renderVcmGateReviewSkillRules } from "../templates/harness/gate-review.js";
 import { renderHarnessEngineerHarnessRules } from "../templates/harness/harness-engineer-agent.js";
 import { renderRootClaudeHarnessRules } from "../templates/harness/claude-root.js";
 import { renderGitignoreHarnessRules } from "../templates/harness/gitignore.js";
@@ -34,6 +34,8 @@ const CLI_DIR = path.dirname(fileURLToPath(import.meta.url));
 const APP_ROOT = path.resolve(CLI_DIR, "../../..");
 const VCM_PACKAGE_VERSION = readVcmPackageVersion(APP_ROOT);
 const MANIFEST_PATH = ".ai/vcm-harness-manifest.json";
+const LEGACY_REVIEWER_AGENT_PATH = ".claude/agents/gate-reviewer.md";
+const REVIEWER_AGENT_PATH = ".claude/agents/reviewer.md";
 const HTML_BLOCK_PATTERN = /<!-- VCM:BEGIN(?:\s+version=\d+)? -->[\s\S]*?<!-- VCM:END -->/m;
 const HASH_BLOCK_PATTERN = /# VCM:BEGIN(?:\s+version=\d+)?\n[\s\S]*?# VCM:END/m;
 const LEGACY_CODEX_HARNESS_PATHS = [
@@ -71,7 +73,7 @@ const AGENT_FRONTMATTER = {
     tester: {
         description: "VCM testing role for validation, test adequacy, approved-scope validation, and risk findings."
     },
-    "gate-reviewer": {
+    reviewer: {
         description: "VCM independent gate review role for architecture plans, validation adequacy, and code diffs.",
         tools: "Read, Grep, Glob, Bash, Write"
     },
@@ -171,13 +173,13 @@ const MANAGED_FILES = [
         content: renderPullRequestTemplateHarnessRules()
     },
     {
-        path: ".claude/agents/gate-reviewer.md",
-        title: "Gate Reviewer Agent",
-        agentName: "gate-reviewer",
+        path: ".claude/agents/reviewer.md",
+        title: "Reviewer Agent",
+        agentName: "reviewer",
         commentStyle: "html",
-        category: "gate-reviewer-agent",
+        category: "reviewer-agent",
         memoryBlock: true,
-        content: renderGateReviewerAgentRules()
+        content: renderReviewerAgentRules()
     },
     {
         path: ".claude/agents/translator.md",
@@ -380,6 +382,7 @@ async function main() {
     const dryRun = args.dryRun;
     const operations = [];
     await assertDirectory(projectRoot, "Project root");
+    await migrateReviewerAgent({ projectRoot, dryRun, operations });
     const manifest = await buildManifest(projectRoot);
     for (const definition of MANAGED_FILES) {
         await installManagedFile({ projectRoot, definition, dryRun, operations });
@@ -405,6 +408,32 @@ async function main() {
     });
     printReport({ projectRoot, dryRun, operations });
 }
+async function migrateReviewerAgent({ projectRoot, dryRun, operations }) {
+    const legacyPath = resolveInside(projectRoot, LEGACY_REVIEWER_AGENT_PATH);
+    const legacyContent = await readOptionalText(legacyPath);
+    if (legacyContent === undefined) {
+        return;
+    }
+    const reviewerPath = resolveInside(projectRoot, REVIEWER_AGENT_PATH);
+    const reviewerContent = await readOptionalText(reviewerPath);
+    if (dryRun) {
+        operations.push(plan(LEGACY_REVIEWER_AGENT_PATH, reviewerContent === undefined
+            ? `rename to ${REVIEWER_AGENT_PATH}`
+            : `delete after ${REVIEWER_AGENT_PATH} was installed`));
+        return;
+    }
+    if (reviewerContent === undefined) {
+        const migratedContent = legacyContent
+            .replace(/^name:[ \t]*gate-reviewer[ \t]*$/m, "name: reviewer")
+            .replace(/^# Gate Reviewer Agent[ \t]*$/m, "# Reviewer Agent");
+        await fs.mkdir(path.dirname(reviewerPath), { recursive: true });
+        await fs.writeFile(reviewerPath, migratedContent, "utf8");
+    }
+    await fs.rm(legacyPath, { force: true });
+    operations.push(done(LEGACY_REVIEWER_AGENT_PATH, reviewerContent === undefined
+        ? `renamed to ${REVIEWER_AGENT_PATH}`
+        : `deleted after ${REVIEWER_AGENT_PATH} was installed`));
+}
 function parseArgs(argv) {
     const args = {
         dryRun: false,

package/dist/backend/services/auto-memory-service.js

@@ -14,7 +14,7 @@ const MEMORY_FILE_DEFINITIONS = [
     { path: ".claude/agents/architect.md", title: "Architect Memory", role: "architect" },
     { path: ".claude/agents/coder.md", title: "Coder Memory", role: "coder" },
     { path: ".claude/agents/tester.md", title: "Tester Memory", role: "tester" },
-    { path: ".claude/agents/gate-reviewer.md", title: "Gate Reviewer Memory", role: "gate-reviewer" },
+    { path: ".claude/agents/reviewer.md", title: "Reviewer Memory", role: "reviewer" },
     { path: ".claude/agents/harness-engineer.md", title: "Harness Engineer Memory", role: "harness-engineer" }
 ];
 export function createAutoMemoryService(deps) {
@@ -151,7 +151,7 @@ export function createAutoMemoryService(deps) {
         const gateSettings = await deps.appSettings.getGateReviewSettings(input.baseRepoRoot, input.taskSlug);
         const roles = ["project-manager", "architect", "coder", "tester"];
         if (gateSettings.enabled) {
-            roles.push("gate-reviewer");
+            roles.push("reviewer");
         }
         const timestamp = now();
         const runId = createRunId(timestamp, "auto");

package/dist/backend/services/claude-hook-service.js

@@ -1,4 +1,4 @@
-import { isGateReviewerRoleName, isHarnessEngineerToolRoleName, isTranslatorToolRoleName, isVcmRoleName } from "../../shared/constants.js";
+import { isHarnessEngineerToolRoleName, isReviewerRoleName, isTranslatorToolRoleName, isVcmRoleName } from "../../shared/constants.js";
 import { VcmError } from "../errors.js";
 import { readTranscriptTurnEvidence } from "./claude-transcript-reply.js";
 import { submitTerminalInput } from "../runtime/terminal-submit.js";
@@ -298,7 +298,7 @@ export function createClaudeHookService(deps) {
             }
         }
         return recordTurnEnd(input, context, eventName, {
-            dispatchRouteFiles: !isGateReviewerRoleName(input.role),
+            dispatchRouteFiles: !isReviewerRoleName(input.role),
             notifyGateway: true,
             settleGuard: true
         });
@@ -347,7 +347,7 @@ export function createClaudeHookService(deps) {
         if (hasCompletionEvidence) {
             await clearStopFailureRecoveryState(context, input.role);
             return recordTurnEnd(input, context, eventName, {
-                dispatchRouteFiles: !isGateReviewerRoleName(input.role),
+                dispatchRouteFiles: !isReviewerRoleName(input.role),
                 notifyGateway: false,
                 settleGuard: true
             });

package/dist/backend/services/gate-review-service.js

@@ -6,11 +6,11 @@ import { VcmError } from "../errors.js";
 import { resolveRepoPath } from "../adapters/filesystem.js";
 import { submitTerminalInput } from "../runtime/terminal-submit.js";
 import { getTaskRuntimeRepoRoot } from "./task-service.js";
-const GATE_REVIEW_AGENT_PATH = ".claude/agents/gate-reviewer.md";
+const REVIEWER_AGENT_PATH = ".claude/agents/reviewer.md";
 const GATE_REVIEW_DIR = ".ai/vcm/gate-reviews";
 const REQUESTS_DIR = ".ai/vcm/gate-reviews/requests";
 const GATE_REVIEW_VERSION = 1;
-const GATE_REVIEWER_ROLE = "gate-reviewer";
+const REVIEWER_ROLE = "reviewer";
 const DEFAULT_REPORT_POLL_INTERVAL_MS = 1000;
 const DEFAULT_REPORT_TIMEOUT_MS = 30 * 60 * 1000;
 const activeRuns = new Set();
@@ -33,7 +33,9 @@ const VALIDATION_ANALYSIS_FIELDS = [
     "Changed Behavior And Risk",
     "Coverage Mapping",
     "Baseline Coverage",
-    "Integration And E2E Coverage",
+    "L2 Integration Coverage",
+    "L3 Trigger Assessment",
+    "L3 End-To-End Coverage",
     "Boundary And Failure Coverage",
     "Public Contract Coverage",
     "Test Integrity",
@@ -431,7 +433,7 @@ export function createGateReviewService(deps) {
             });
             await updateRequestStatus(deps.fs, context, requestId, "running", { startedAt: timestamp });
             const reviewDir = resolveRepoPath(context.taskRepoRoot, GATE_REVIEW_DIR);
-            const agentPath = resolveRepoPath(context.repoRoot, GATE_REVIEW_AGENT_PATH);
+            const agentPath = resolveRepoPath(context.repoRoot, REVIEWER_AGENT_PATH);
             const prompt = buildGatePrompt(context, gate, requestId, codeDiffInput, codeDiffSources);
             await deps.fs.ensureDir(reviewDir);
             await deps.fs.ensureDir(resolveRepoPath(context.taskRepoRoot, REQUESTS_DIR));
@@ -439,20 +441,20 @@ export function createGateReviewService(deps) {
             if (!(await deps.fs.pathExists(agentPath))) {
                 throw new VcmError({
                     code: "GATE_REVIEW_AGENT_MISSING",
-                    message: `${GATE_REVIEW_AGENT_PATH} does not exist.`,
+                    message: `${REVIEWER_AGENT_PATH} does not exist.`,
                     statusCode: 409,
                     hint: "Apply the VCM harness before requesting Gate Review Gates."
                 });
             }
-            const session = await ensureGateReviewerSession(context);
+            const session = await ensureReviewerSession(context);
             await submitTerminalInput(deps.runtime, session.id, prompt);
-            await deps.sessionService.markRoleActivityRunning(context.repoRoot, context.taskSlug, GATE_REVIEWER_ROLE, session.id);
+            await deps.sessionService.markRoleActivityRunning(context.repoRoot, context.taskSlug, REVIEWER_ROLE, session.id);
             await deps.roundService.recordRoleTurnEvent({
                 repoRoot: context.repoRoot,
                 stateRepoRoot: context.taskRepoRoot,
                 stateRoot: context.stateRoot,
                 taskSlug: context.taskSlug,
-                role: GATE_REVIEWER_ROLE,
+                role: REVIEWER_ROLE,
                 eventName: "UserPromptSubmit"
             });
             const parsed = await waitForGateReport(deps.fs, context.taskRepoRoot, gate, requestId, now(), {
@@ -501,24 +503,24 @@ export function createGateReviewService(deps) {
             activeRuns.delete(runKey);
         }
     }
-    async function ensureGateReviewerSession(context) {
-        const existing = await deps.sessionService.getRoleSession(context.repoRoot, context.taskSlug, GATE_REVIEWER_ROLE);
+    async function ensureReviewerSession(context) {
+        const existing = await deps.sessionService.getRoleSession(context.repoRoot, context.taskSlug, REVIEWER_ROLE);
         if (existing?.status === "running" && deps.runtime.getSession(existing.id)) {
             return existing;
         }
         if (existing?.claudeSessionId) {
             try {
-                return await deps.sessionService.resumeRoleSession(context.repoRoot, context.taskSlug, GATE_REVIEWER_ROLE, {
+                return await deps.sessionService.resumeRoleSession(context.repoRoot, context.taskSlug, REVIEWER_ROLE, {
                     cols: 100,
                     rows: 28,
                     model: "default"
                 });
             }
             catch {
-                // Fall through to a fresh Gate Reviewer terminal if the saved session cannot be resumed.
+                // Fall through to a fresh Reviewer terminal if the saved session cannot be resumed.
             }
         }
-        return deps.sessionService.startRoleSession(context.repoRoot, context.taskSlug, GATE_REVIEWER_ROLE, {
+        return deps.sessionService.startRoleSession(context.repoRoot, context.taskSlug, REVIEWER_ROLE, {
             cols: 100,
             rows: 28,
             model: "default"
@@ -891,7 +893,7 @@ async function computeInputHash(deps, taskRepoRoot, gate, codeDiffInput, codeDif
         "CLAUDE.md",
         ".claude/agents/architect.md",
         ".claude/agents/coder.md",
-        ".claude/agents/gate-reviewer.md",
+        ".claude/agents/reviewer.md",
         ".claude/agents/tester.md",
         ".claude/skills/vcm-gate-review/SKILL.md",
         ".ai/tools/request-gate-review",
@@ -1048,13 +1050,13 @@ function buildGatePrompt(context, gate, requestId, codeDiffInput, codeDiffSource
         ? "\nDiff: inspect git status/diff in Worktree."
         : "";
     const architectureContract = gate === "architecture-plan"
-        ? "\n\nComplete every Architecture Analysis field required by the Gate Reviewer role with concrete current-worktree evidence before deciding."
+        ? "\n\nComplete every Architecture Analysis field required by the Reviewer role with concrete current-worktree evidence before deciding."
         : "";
     const validationContract = gate === "validation-adequacy"
-        ? "\n\nComplete every Validation Analysis field required by the Gate Reviewer role with concrete current-worktree production and test evidence before deciding."
+        ? "\n\nComplete every Validation Analysis field required by the Reviewer role with concrete current-worktree production and test evidence before deciding."
         : "";
     const codeDiffContract = gate === "code-diff"
-        ? "\n\nComplete every Code Diff Analysis field required by the Gate Reviewer role with concrete evidence from the named commit range before deciding."
+        ? "\n\nComplete every Code Diff Analysis field required by the Reviewer role with concrete evidence from the named commit range before deciding."
         : "";
     const codeDiffSection = gate === "code-diff" && codeDiffInput
         ? `
@@ -1109,7 +1111,7 @@ async function waitForGateReport(fs, taskRepoRoot, gate, requestId, timestamp, o
     const detail = errorMessage(lastError);
     throw new VcmError({
         code: "GATE_REVIEW_REPORT_TIMEOUT",
-        message: `Gate Reviewer did not produce a valid ${gate} report within ${Math.round(options.timeoutMs / 1000)}s.`,
+        message: `Reviewer did not produce a valid ${gate} report within ${Math.round(options.timeoutMs / 1000)}s.`,
         statusCode: 504,
         hint: detail
     });

package/dist/backend/services/harness-service.js

@@ -5,7 +5,7 @@ import { renderArchitectHarnessRules } from "../templates/harness/architect-agen
 import { renderCoderHarnessRules } from "../templates/harness/coder-agent.js";
 import { renderCoderWorkerHarnessRules } from "../templates/harness/coder-worker-agent.js";
 import { renderArchitectScaffoldWorkerHarnessRules } from "../templates/harness/architect-scaffold-worker-agent.js";
-import { renderGateReviewerAgentRules, renderRequestGateReviewTool, renderTranslatorAgentRules, renderVcmGateReviewSkillRules } from "../templates/harness/gate-review.js";
+import { renderReviewerAgentRules, renderRequestGateReviewTool, renderTranslatorAgentRules, renderVcmGateReviewSkillRules } from "../templates/harness/gate-review.js";
 import { renderHarnessEngineerHarnessRules } from "../templates/harness/harness-engineer-agent.js";
 import { renderRootClaudeHarnessRules } from "../templates/harness/claude-root.js";
 import { renderGitignoreHarnessRules } from "../templates/harness/gitignore.js";
@@ -183,12 +183,12 @@ const HARNESS_FILES = [
         renderRules: renderRestartArchitectSkillRules
     },
     {
-        kind: "agent-gate-reviewer",
-        path: ".claude/agents/gate-reviewer.md",
-        title: "Gate Reviewer Agent",
+        kind: "agent-reviewer",
+        path: ".claude/agents/reviewer.md",
+        title: "Reviewer Agent",
         memoryBlock: true,
-        frontmatter: renderAgentFrontmatter("gate-reviewer", "VCM independent gate review role for architecture plans, validation adequacy, and code diffs.", { tools: "Read, Grep, Glob, Bash, Write" }),
-        renderRules: renderGateReviewerAgentRules
+        frontmatter: renderAgentFrontmatter("reviewer", "VCM independent gate review role for architecture plans, validation adequacy, and code diffs.", { tools: "Read, Grep, Glob, Bash, Write" }),
+        renderRules: renderReviewerAgentRules
     },
     {
         kind: "agent-translator",
@@ -1279,7 +1279,7 @@ async function analyzeLegacyCodexHarnessPaths(fs, repoRoot) {
         changes.push({
             path: relativePath,
             action: "delete",
-            reason: "Legacy Codex harness path is obsolete; VCM now uses Claude Code Gate Reviewer and Translator roles."
+            reason: "Legacy Codex harness path is obsolete; VCM now uses Claude Code Reviewer and Translator roles."
         });
     }
     return changes;

package/dist/backend/services/task-launch-service.js

@@ -1,4 +1,4 @@
-import { CORE_VCM_ROLE_DEFINITIONS, GATE_REVIEWER_ROLE_DEFINITION, VCM_ROLE_NAMES } from "../../shared/constants.js";
+import { CORE_VCM_ROLE_DEFINITIONS, REVIEWER_ROLE_DEFINITION, VCM_ROLE_NAMES } from "../../shared/constants.js";
 import { VcmError } from "../errors.js";
 import { getTaskRuntimeRepoRoot } from "./task-service.js";
 const ONE_CLICK_SESSION_COLS = 100;
@@ -25,10 +25,10 @@ export function createTaskLaunchService(deps) {
             mode
         });
     }
-    function composeRoleDefinitions(gateReviewerEnabled) {
+    function composeRoleDefinitions(reviewerEnabled) {
         return [
             ...CORE_VCM_ROLE_DEFINITIONS,
-            ...(gateReviewerEnabled ? [GATE_REVIEWER_ROLE_DEFINITION] : [])
+            ...(reviewerEnabled ? [REVIEWER_ROLE_DEFINITION] : [])
         ];
     }
     // Skip a running role, resume one that has a prior Claude session, otherwise

package/dist/backend/templates/handoff.js

@@ -157,6 +157,28 @@ TBD
 
 TBD
 
+## L3 Coverage
+
+L3 Required: yes|no
+
+### Trigger Assessment
+
+TBD
+
+### Affected End-To-End Flows
+
+| Flow | Trigger | Case ID | Test File | Entry Point | Final Observable Result | Action | Result |
+| --- | --- | --- | --- | --- | --- | --- | --- |
+| TBD | TBD | TBD | TBD | TBD | TBD | TBD | TBD |
+
+### L3 Commands And Evidence
+
+TBD
+
+### Not-Required Evidence
+
+TBD
+
 ## Commands Run Or Checked
 
 TBD

package/dist/backend/templates/harness/gate-review.js

@@ -1,10 +1,10 @@
 import { renderRoleMemoryRules } from "./role-memory.js";
-export function renderGateReviewerAgentRules() {
+export function renderReviewerAgentRules() {
     return `## Role
 
-You are VCM \`gate-reviewer\`.
+You are VCM \`reviewer\`.
 
-${renderRoleMemoryRules("gate-reviewer")}
+${renderRoleMemoryRules("reviewer")}
 
 Review only the gate in the VCM prompt. Use the task and worktree paths named there. Project memory may orient you, but only current worktree evidence can decide the gate.
 
@@ -150,6 +150,31 @@ UI, CLI/tooling, hook, session,
 persistence, worktree, external-process, and other important user or system
 paths have integration or E2E coverage that exercises real behavior.
 
+Independently apply the Tester L3 trigger rules to the accepted task, actual
+production changes, public contracts, and documented L3 flows. Do not accept
+\`L3 Required: no\` merely because Tester selected it.
+
+For every affected end-to-end flow:
+
+- Verify that the trigger assessment matches the actual production path.
+- Inspect the actual L3 test file and case.
+- Verify that the case starts from the documented system entry point and reaches
+  the final observable result through the project-owned production path.
+- Verify that internal production wiring is not mocked, replaced, or bypassed.
+- Verify that the assertions directly cover the changed behavior and any changed
+  failure, retry, recovery, or lifecycle path.
+- Verify that an existing case was updated when its prior assertions did not
+  cover the change.
+- Verify the recorded command and result.
+
+Request changes when required L3 coverage is missing, replaced by L2, not
+executed, mapped only to an old case without relevant assertions, bypasses the
+production path, or is classified as unnecessary without satisfying every
+not-required condition.
+
+A required L3 gap is acceptable only when \`test-report.md\` contains the user's
+exact approval for that specific Coverage Gap.
+
 Inspect boundary, failure, cancellation, retry, restart, recovery,
 concurrency, repeated-action, stale-state, cleanup, and compatibility paths
 when they are relevant to the changed behavior. Check that tests were not
@@ -160,9 +185,8 @@ Do not approve only because \`Test Result: pass\` or all recorded commands are
 green. Request changes when the report is incomplete or inconsistent with the
 actual tests, validation level does not match risk, an important behavior has
 no concrete coverage mapping, a required check was skipped, required coverage
-is unavailable, or a current-task coverage gap remains. A concrete risk-based
-reason may show that integration or E2E coverage is unnecessary; unavailable
-required coverage without exact user approval is not an approval reason.
+is unavailable, or a current-task coverage gap remains. Required coverage
+without exact user approval is not an approval reason.
 
 Treat every unresolved required-coverage item as gate-blocking unless
 \`test-report.md\` contains the user's exact approval routed by project-manager.
@@ -278,7 +302,9 @@ Use this findings structure:
 - Changed Behavior And Risk:
 - Coverage Mapping:
 - Baseline Coverage:
-- Integration And E2E Coverage:
+- L2 Integration Coverage:
+- L3 Trigger Assessment:
+- L3 End-To-End Coverage:
 - Boundary And Failure Coverage:
 - Public Contract Coverage:
 - Test Integrity:
@@ -339,7 +365,9 @@ If there are no findings, write:
 - Changed Behavior And Risk:
 - Coverage Mapping:
 - Baseline Coverage:
-- Integration And E2E Coverage:
+- L2 Integration Coverage:
+- L3 Trigger Assessment:
+- L3 End-To-End Coverage:
 - Boundary And Failure Coverage:
 - Public Contract Coverage:
 - Test Integrity:
@@ -459,7 +487,7 @@ When VCM sends \`[VCM GATE REVIEW CALLBACK]\`, read the named report path.
 - \`failed\`: stop and ask the user to retry, skip, or override in VCM.
 - \`skipped\` or \`overridden\`: record the exception reason in PM context and continue only as appropriate.
 
-Do not ask Gate Reviewer to choose owners, fixes, Replan, or user-intervention needs. PM routes those decisions through normal VCM responsibilities.`;
+Do not ask Reviewer to choose owners, fixes, Replan, or user-intervention needs. PM routes those decisions through normal VCM responsibilities.`;
 }
 export function renderRequestGateReviewTool() {
     return `#!/usr/bin/env python3
@@ -691,7 +719,7 @@ def input_hash(root: Path, gate: str, sources: list[str] | None = None, gate_rec
         "CLAUDE.md",
         ".claude/agents/architect.md",
         ".claude/agents/coder.md",
-        ".claude/agents/gate-reviewer.md",
+        ".claude/agents/reviewer.md",
         ".claude/agents/tester.md",
         ".claude/skills/vcm-gate-review/SKILL.md",
         ".ai/tools/request-gate-review",

package/dist/backend/templates/harness/project-manager-agent.js

@@ -40,7 +40,7 @@ PM Managed Mode applies only when the user explicitly asks to complete the curre
 - PM must drive the accepted task to completion through the normal VCM flow.
 - PM must not reduce, defer, reinterpret, skip, or move requested work outside the current task.
 - PM must not use workload, task size, context size, implementation difficulty, dependency choice, refactor need, testing effort, or number of iterations as a reason to ask the user.
-- Technical execution questions are handled inside the VCM flow. PM routes them to Architect, Coder, Tester, or Gate Reviewer according to role responsibility.
+- Technical execution questions are handled inside the VCM flow. PM routes them to Architect, Coder, Tester, or Reviewer according to role responsibility.
 - PM may defer non-blocking user-facing questions until the final user report, but only when continuing does not require user intent, external authorization, or a user-approved exception.
 - Deferred questions remain part of the current task report. They must not become follow-up scope unless the user explicitly creates a new task.
 - PM must pause and ask the user only when the task cannot proceed without user intent or real-world authorization: unclear or conflicting requested outcome, required account/secret/test environment/data access, real cost, production permission, sensitive data access, destructive or irreversible real-world action, durable-doc conflict requiring user choice, or a required user-approved exception.
@@ -334,7 +334,7 @@ PM may lightly rewrite the user's words to:
 - Do not run \`code-diff\` for incomplete, failed, planning-only, Docs-Only Flow, Validation-Only Flow, PR-Preparation Flow, or Communication-Only Flow.
 - Gate Review trigger points apply only when the active delivery flow reaches that milestone. Do not run Gate Review for Communication-Only Flow.
 - On a callback, accept only \`approve\` or \`request_changes\`. Apply \`request_changes\` through the allowed branch defined by the active flow; in Code-Change Flow use Architecture Plan Revision, Code-Diff Correction, or Validation Revision according to the gate.
-- Do not ask Gate Reviewer to choose owners, fixes, Replan, or user-intervention needs.
+- Do not ask Reviewer to choose owners, fixes, Replan, or user-intervention needs.
 - Record gate decision, report path, and any skip or override reason.
 
 ### Partial Role Results

package/dist/backend/templates/harness/tester-agent.js

@@ -29,9 +29,7 @@ ${renderRoleMemoryRules("tester")}
 - If required L0/L1 coverage is missing or weak, add or update the required tests. If the coverage cannot be completed, return \`Test Result: fail\` with concrete blocking evidence.
 - Own L2/L3/L4 final-validation design, execution, and acceptance evidence.
 - Targeted diagnostic L2 checks run by Coder or Architect are implementation evidence only and do not replace Tester final validation.
-- Choose validation level by risk. Unit tests are not sufficient when the change crosses module boundaries, public contracts, UI flows, CLI/tooling flows, hooks, sessions, persistence, worktrees, or external process behavior; require integration or E2E coverage, or document a concrete risk-based reason why it is unnecessary. Unavailable required coverage is a blocking validation gap.
-- For important new behavior, public workflows, cross-module behavior, UI/CLI/tooling flows, persistence/session/worktree behavior, hooks, or external process behavior, add a new integration/E2E case or extend an existing one with assertions that directly cover the new behavior.
-- Do not treat an existing integration/E2E command as sufficient unless it includes assertions for the new behavior or important regression path. Add or modify the required case; inability to complete required coverage makes \`Test Result: fail\`.
+- Use L2 integration coverage when changed behavior crosses internal module or component boundaries and can be completely proved from a stable integration entry point without triggering the mandatory L3 rules below.
 - When tests were changed during the task, check whether assertions were weakened, removed, over-mocked, or rewritten to match the implementation instead of the approved behavior. Report this as a validation gap unless the approved contract changed.
 - Apply \`docs/CODING_STANDARDS.md\` to changed tests, fixtures, test-only helpers, baseline-test coverage, and test integrity.
 - Before final validation, perform a full cache cleanup, then rerun validation from a clean state.
@@ -51,7 +49,6 @@ ${renderRoleMemoryRules("tester")}
 - Add anti-hardcode coverage when risk warrants it: use non-fixture inputs, boundary values, negative cases, repeated actions, and assertions through public/runtime paths.
 - Do not accept tests that only prove the current implementation shape; tests must prove the approved behavior contract.
 - Treat architect-flagged public contracts, migrations, auth, data flow, routing, or dependency changes as inputs for tester-owned validation design.
-- Record skipped L3 checks in \`.ai/vcm/handoffs/test-report.md\` with the reason.
 - Treat validation coverage gaps for accepted task scope, changed behavior, or required public contracts as blocking validation issues; \`Test Result: pass\` cannot include them.
 - Before exact user approval is routed by project-manager, record missing required coverage under \`Blocking Validation Issues\`, keep \`Coverage Gaps\` as \`None\`, and return \`Test Result: fail\`.
 - Add a Coverage Gap only after project-manager routes the user's exact approval for that specific unresolved gap. Record the approval verbatim in \`User Approval Evidence\`.
@@ -59,6 +56,59 @@ ${renderRoleMemoryRules("tester")}
 - If a required validation check is skipped or cannot complete, \`Test Result\` must be \`fail\`.
 - Update \`docs/TESTING.md\` when validation strategy, commands, level mapping, integration/E2E case definitions, selection rules, final-validation cleanup, test gaps, or test expectations change.
 
+### Mandatory L3 End-To-End Coverage
+
+L3 validates a complete externally observable flow from a project-defined
+system entry point, through the actual project-owned production path, to its
+final observable result.
+
+Do not mock, replace, or bypass the project-owned production path being
+validated. External dependencies may use controlled substitutes only when
+allowed by \`docs/TESTING.md\`.
+
+L3 is required when any of the following is true:
+
+- The accepted task adds a new externally reachable end-to-end flow.
+- The task changes the input, output, error result, persisted result, external
+  side effect, or other observable behavior of an end-to-end flow.
+- The changed production path is covered by an existing L3 case in
+  \`docs/TESTING.md\`.
+- The task changes completion, failure, cancellation, retry, recovery, timeout,
+  idempotency, duplicate-event, or out-of-order behavior that affects the final
+  result of an end-to-end flow.
+- The task changes a public API, event, message, storage, migration, or other
+  external contract used by an end-to-end flow.
+- The task changes a cross-component critical invariant that can be proved only
+  through the complete production path.
+- The task fixes a defect that passed L1/L2 but occurred in an integrated,
+  staging, production, or other complete-system flow.
+
+L3 is not required only when all of the following are true:
+
+- No externally observable end-to-end behavior is added or changed.
+- No production path covered by a documented L3 case is affected.
+- No end-to-end lifecycle, external contract, or critical invariant is changed.
+- L1 or L2 can completely prove the accepted behavior from a stable test entry
+  point.
+
+Task size, changed-file count, implementation size, existing unit tests, or a
+green L2 result are not reasons to skip required L3 coverage.
+
+For every affected end-to-end flow:
+
+- Run an existing L3 case when its assertions already cover the changed behavior.
+- Update an existing L3 case when the flow is covered but the changed behavior
+  is not asserted.
+- Add a new L3 case when the task creates a new flow or no existing case covers
+  it.
+- Add or update assertions for any failure, retry, recovery, or lifecycle path
+  changed by the task.
+
+Required L3 coverage cannot be replaced by L2. If the required case cannot be
+added or executed, return \`Test Result: fail\` and record the missing coverage as
+a blocking validation issue unless the user has explicitly approved that exact
+Coverage Gap.
+
 ### Testing Documentation
 
 - Own \`docs/TESTING.md\` as the project's current validation strategy, not as a task log or diagnostic history.
@@ -76,6 +126,27 @@ ${renderRoleMemoryRules("tester")}
 ### Outputs
 
 - Write \`.ai/vcm/handoffs/test-report.md\` with \`Test Result: pass|fail\`, evidence reviewed, tests added or updated, coverage mapping, commands run or checked, validation results, failed expectations, reproduction steps, skipped checks with reasons, coverage gaps, blocking validation issues, and user approval evidence.
+- \`test-report.md\` must include this L3 section:
+
+\`\`\`md
+## L3 Coverage
+
+L3 Required: yes|no
+
+### Trigger Assessment
+
+### Affected End-To-End Flows
+
+| Flow | Trigger | Case ID | Test File | Entry Point | Final Observable Result | Action | Result |
+| --- | --- | --- | --- | --- | --- | --- | --- |
+
+### L3 Commands And Evidence
+
+### Not-Required Evidence
+\`\`\`
+
+- When \`L3 Required: yes\`, include at least one complete flow-to-case mapping. \`Action\` must be \`run-existing\`, \`updated\`, or \`added\`.
+- When \`L3 Required: no\`, use \`Not-Required Evidence\` to prove every condition in the L3 not-required rule.
 - In Validation-Only Flow, if tests, fixtures, test-only helpers, or \`docs/TESTING.md\` changed, commit those changes before reporting and record the changed files and commit in \`test-report.md\`. If no tracked files changed, record that no commit was required.
 - \`test-report.md\` is the current validation evidence, not a log; when rewriting it, carry forward still-unresolved findings or explicitly mark them resolved instead of dropping them.
 - In \`Coverage Mapping\`, map each accepted changed behavior or relevant risk to its validation level, actual test file and case or external evidence, exercised entry path and key assertions, result, and any remaining gap.
@@ -85,7 +156,7 @@ ${renderRoleMemoryRules("tester")}
 - When \`Test Result: fail\`, \`Blocking Validation Issues\` must list concrete blocking evidence.
 - When \`Coverage Gaps\` is not \`None\`, \`Test Result\` must be \`fail\`, \`User Approval Evidence\` must contain the user's exact authorization, and every recorded gap must match that authorization.
 - When no gap has been approved, \`User Approval Evidence\` must be \`None\`.
-- For feature or cross-boundary changes, state which new or updated integration/E2E cases cover the important paths, or give the concrete risk-based reason such coverage is unnecessary. If required coverage is unavailable, report it as a blocking issue.
+- For feature or cross-boundary changes, map required L2 integration coverage and mandatory L3 coverage separately. If required coverage is unavailable, report it as a blocking issue.
 - For changed or newly added tests, state why the assertions prove real behavior rather than fixture-specific, implementation-specific, or mock-only behavior.
 - Report confirmed unresolved issues that should survive current-task cleanup in \`.ai/vcm/handoffs/test-report.md\`; do not write \`.ai/vcm/handoffs/known-issues.md\` (architect-owned).