@aibridge/cli 0.4.0 → 0.5.0

package/README.md

@@ -28,7 +28,7 @@ npx skills add ycmjason/aibridge
 That's it. The skill runs the CLI on demand via `npx -y @aibridge/cli` — nothing else to install. Ask your agent to "use aibridge", or try it yourself:
 
 ```bash
-npx -y @aibridge/cli subagent --model xai-grok/grok-4.5 "summarize the architecture of this repo"
+npx -y @aibridge/cli subagent --model xai-grok/grok-4.6 "summarize the architecture of this repo"
 ```
 
 <sup>Want the `aibridge` command on your PATH for manual use? `npm i -g @aibridge/cli` (optional).</sup>
@@ -37,10 +37,10 @@ npx -y @aibridge/cli subagent --model xai-grok/grok-4.5 "summarize the architect
 
 | Command | Use when |
 |---|---|
-| `aibridge plan --model xai-grok/grok-4.5 --out plan.md "<task>"` | You want a delegate model to study the repo and expand a task into a detailed, reviewable **plan file** before any code is written |
-| `aibridge implement --model google-antigravity/gemini-3.6-flash <plan.md>` | You have an approved plan file and want it executed in place — with your project's **real typecheck and tests** run until green |
-| `aibridge review --model xai-grok/grok-4.5 --out review.md [--plan <plan.md>]` | You want a **different model** to pressure-test the working-tree diff against the plan contract (over-reach is a finding) — or to review the plan itself before implementing |
-| `aibridge subagent --model xai-grok/grok-4.5 "<task>"` | A self-contained task deserves a concurrent delegate, a cross-model second opinion, or a red-team pass |
+| `aibridge plan --model xai-grok/grok-4.6 --out plan.md "<task>"` | You want a delegate model to study the repo and expand a task into a detailed, reviewable **plan file** before any code is written |
+| `aibridge implement --model google-antigravity/gemini-3.7-flash <plan.md>` | You have an approved plan file and want it executed in place — with your project's **real typecheck and tests** run until green |
+| `aibridge review --model xai-grok/grok-4.6 --out review.md [--plan <plan.md>]` | You want a **different model** to pressure-test the working-tree diff against the plan contract (over-reach is a finding) — or to review the plan itself before implementing |
+| `aibridge subagent --model xai-grok/grok-4.6 "<task>"` | A self-contained task deserves a concurrent delegate, a cross-model second opinion, or a red-team pass |
 | `aibridge image-gen --model openai-codex/gpt-5.6-sol --out out.png "<prompt>"` | You need a real raster image — on a Codex, Antigravity, or Grok seat, with render verification |
 | `aibridge models [--json]` | You need the exact facts for every registered model seat (accepted efforts, image format, pinned model ID) |
 | `aibridge quota` | Two-second check of every backend's remaining quota before you pipeline work |
@@ -49,10 +49,10 @@ npx -y @aibridge/cli subagent --model xai-grok/grok-4.5 "summarize the architect
 The three verbs compose into an orchestrator-driven loop your agent stays in charge of:
 
 ```
-aibridge plan --model xai-grok/grok-4.5 --out plan.md "add rate limiting to the API"   # delegate writes plan.md
+aibridge plan --model xai-grok/grok-4.6 --out plan.md "add rate limiting to the API"   # delegate writes plan.md
 # → your agent reads, edits, approves the plan
-aibridge implement --model google-antigravity/gemini-3.6-flash plan.md                 # another model executes it, runs your gates
-aibridge review --model xai-grok/grok-4.5 --out review.md --plan plan.md               # a third seat cross-checks the diff
+aibridge implement --model google-antigravity/gemini-3.7-flash plan.md                 # another model executes it, runs your gates
+aibridge review --model xai-grok/grok-4.6 --out review.md --plan plan.md               # a third seat cross-checks the diff
 ```
 
 Plan files — not their contents — travel between stages, so the loop is nearly free on your agent's context.
@@ -62,7 +62,7 @@ Plan files — not their contents — travel between stages, so the loop is near
 - **The skill carries judgment; the CLI owns execution.** The skill teaches your agent prompt-craft, seat selection, and when to gate; the CLI deterministically drives the backing CLIs, captures their output, verifies results (a "generated image" under 100 KB is a code-drawn fake, an empty answer is a quota death), and logs every run.
 - **Seats stay cross-model by default.** Grok plans and reviews, Gemini implements — a model never reviews its own diff, and independent eyes catch what shared blind spots miss.
 - **No API keys.** Delegation runs on the backing CLIs' existing logins, each spending its own quota. (The skill treats a backend that shares your agent's own quota pool as a last resort.)
-- **Models are canonical slugs**: `<vendor>-<cli>/<model>[-<effort>]` — e.g. `xai-grok/grok-4.5`, `google-antigravity/gemini-3.6-flash`, `openai-codex/gpt-5.6-sol-high`, `anthropic-claude/opus-5`. No aliases — not short ones, and not moving vendor aliases like `opus`: every seat pins an exact model version. `aibridge <command> --help` lists every seat.
+- **Models are canonical slugs**: `<vendor>-<cli>/<model>[-<effort>]` — e.g. `xai-grok/grok-4.6`, `google-antigravity/gemini-3.7-flash`, `openai-codex/gpt-5.6-sol-high`, `anthropic-claude/opus-5`. No aliases — not short ones, and not moving vendor aliases like `opus`: every seat pins an exact model version. `aibridge <command> --help` lists every seat.
 
 ## Tell your agent when to reach for it
 

package/dist/cli.mjs

@@ -1,5 +1,5 @@
 #!/usr/bin/env node
-import { r as runCli, t as buildContext } from "./context-v_ns4n_W.mjs";
+import { r as runCli, t as buildContext } from "./context-F8WLXzPv.mjs";
 //#region src/cli.ts
 await runCli(buildContext(process), process.argv.slice(2));
 //#endregion

package/dist/{context-v_ns4n_W.mjs → context-F8WLXzPv.mjs}

@@ -11,10 +11,21 @@ import * as codex from "@aibridge/driver-codex";
 import { fetchCodexQuota } from "@aibridge/driver-codex";
 import * as grok from "@aibridge/driver-grok";
 import { fetchGrokQuota } from "@aibridge/driver-grok";
-import { runCaptured } from "@aibridge/proc";
+import { isAuthExpired, runCaptured } from "@aibridge/proc";
 import { randomBytes } from "node:crypto";
 //#region src/models.ts
 const MODELS = {
+	"xai-grok/grok-4.6": {
+		slug: "xai-grok/grok-4.6",
+		backend: "grok",
+		backendModel: "grok-4.6",
+		efforts: [
+			"low",
+			"medium",
+			"high"
+		],
+		brief: "xAI Grok 4.6 via grok CLI — own xAI login; ~30 req/min, ~1k msgs/day, single-flight"
+	},
 	"xai-grok/grok-4.5": {
 		slug: "xai-grok/grok-4.5",
 		backend: "grok",
@@ -26,6 +37,18 @@ const MODELS = {
 		],
 		brief: "xAI Grok 4.5 via grok CLI — own xAI login; ~30 req/min, ~1k msgs/day, single-flight"
 	},
+	"google-antigravity/gemini-3.7-flash": {
+		slug: "google-antigravity/gemini-3.7-flash",
+		backend: "agy",
+		backendModel: "gemini-3.7-flash",
+		efforts: [
+			"low",
+			"medium",
+			"high"
+		],
+		defaultEffort: "high",
+		brief: "Google Gemini 3.7 Flash via agy — own Antigravity login; quota shared across all Gemini tiers (not yet itemised per-model, so no exhaustion preflight)"
+	},
 	"google-antigravity/gemini-3.6-flash": {
 		slug: "google-antigravity/gemini-3.6-flash",
 		backend: "agy",
@@ -515,6 +538,7 @@ function evaluateAgyPreflight(snapshot, backendModel) {
 		}
 		return {
 			ok: false,
+			kind: "quota",
 			message: `agy model "${backendModel}" is quota-exhausted`,
 			resetAt
 		};
@@ -524,12 +548,14 @@ function evaluateAgyPreflight(snapshot, backendModel) {
 function evaluateCodexPreflight(snapshot) {
 	if (snapshot.limitReached) return {
 		ok: false,
+		kind: "quota",
 		message: "codex quota limit reached",
 		resetAt: snapshot.windows.find((w) => w.resetAt)?.resetAt
 	};
 	const exhaustedWindow = snapshot.windows.find((w) => w.usedPercent >= 100);
 	if (exhaustedWindow) return {
 		ok: false,
+		kind: "quota",
 		message: "codex quota limit reached",
 		resetAt: exhaustedWindow.resetAt
 	};
@@ -538,6 +564,7 @@ function evaluateCodexPreflight(snapshot) {
 function evaluateGrokPreflight(snapshot) {
 	if (snapshot.usedPercent !== void 0 && snapshot.usedPercent >= 100) return {
 		ok: false,
+		kind: "quota",
 		message: "grok credit quota exhausted",
 		resetAt: snapshot.periodEnd
 	};
@@ -550,6 +577,12 @@ async function preflightModel(resolved) {
 	try {
 		return evaluateAgyPreflight(await fetchAgyQuota(), backendModelId(resolved));
 	} catch (err) {
+		if (isAuthExpired(err)) return {
+			ok: false,
+			kind: "auth",
+			message: err.message,
+			resetAt: void 0
+		};
 		return {
 			ok: true,
 			warning: `quota preflight failed (${err.message}); proceeding`
@@ -560,6 +593,12 @@ async function preflightCodex() {
 	try {
 		return evaluateCodexPreflight(await fetchCodexQuota());
 	} catch (err) {
+		if (isAuthExpired(err)) return {
+			ok: false,
+			kind: "auth",
+			message: err.message,
+			resetAt: void 0
+		};
 		return {
 			ok: true,
 			warning: `quota preflight failed (${err.message}); proceeding`
@@ -570,6 +609,12 @@ async function preflightGrok() {
 	try {
 		return evaluateGrokPreflight(await fetchGrokQuota());
 	} catch (err) {
+		if (isAuthExpired(err)) return {
+			ok: false,
+			kind: "auth",
+			message: err.message,
+			resetAt: void 0
+		};
 		return {
 			ok: true,
 			warning: `quota preflight failed (${err.message}); proceeding`
@@ -586,6 +631,7 @@ function formatReset$1(resetTime) {
 	return `${new Date(resetTime).toLocaleTimeString()} (in ${rel})`;
 }
 function renderPreflightRefusal(cmd, verdict) {
+	if (verdict.kind === "auth") return `aibridge ${cmd}: refusing — ${verdict.message}. Running with --no-preflight would only send the delegate in unauthenticated. Or use a different --model.`;
 	const resetClause = verdict.resetAt ? ` Resets ${formatReset$1(verdict.resetAt)}.` : "";
 	return `aibridge ${cmd}: refusing — ${verdict.message}.${resetClause} Use --no-preflight to override, or a claude-backend fallback (subagent --model sonnet|opus — bills the Claude subscription).`;
 }
@@ -1537,7 +1583,7 @@ const fullDescription = [
 	"",
 	"Available models (canonical slug):",
 	...listModelHelpLines(),
-	"Recommended first choice: xai-grok/grok-4.5. Whichever seat runs on the same provider as the",
+	"Recommended first choice: xai-grok/grok-4.6. Whichever seat runs on the same provider as the",
 	"agent you orchestrate from is your last resort — it spends the pool you are already burning."
 ].join("\n");
 const subagent = buildCommand({

package/dist/index.d.mts

@@ -165,6 +165,7 @@ type PreflightVerdict = {
   readonly warning?: string;
 } | {
   readonly ok: false;
+  readonly kind: 'auth' | 'quota';
   readonly message: string;
   readonly resetAt: string | undefined;
 };
@@ -173,6 +174,7 @@ declare function evaluateCodexPreflight(snapshot: CodexQuotaSnapshot): Preflight
 declare function preflightModel(resolved: ResolvedModel): Promise<PreflightVerdict>;
 declare function preflightCodex(): Promise<PreflightVerdict>;
 declare function renderPreflightRefusal(cmd: string, verdict: {
+  kind: 'auth' | 'quota';
   message: string;
   resetAt: string | undefined;
 }): string;

package/dist/index.mjs

@@ -1,2 +1,2 @@
-import { S as supportsImageGen, _ as backendModelId, a as readRunLogs, b as listModelHelpLines, c as evaluateCodexPreflight, d as renderPreflightRefusal, f as delegate, g as MODELS, h as positiveIntSeconds, i as listRuns, l as preflightCodex, m as nonEmptyPrompt, n as app, o as startRun, p as getDriver, r as runCli, s as evaluateAgyPreflight, t as buildContext, u as preflightModel, v as formatImageGenModelError, x as resolveModel, y as formatUnknownModelError } from "./context-v_ns4n_W.mjs";
+import { S as supportsImageGen, _ as backendModelId, a as readRunLogs, b as listModelHelpLines, c as evaluateCodexPreflight, d as renderPreflightRefusal, f as delegate, g as MODELS, h as positiveIntSeconds, i as listRuns, l as preflightCodex, m as nonEmptyPrompt, n as app, o as startRun, p as getDriver, r as runCli, s as evaluateAgyPreflight, t as buildContext, u as preflightModel, v as formatImageGenModelError, x as resolveModel, y as formatUnknownModelError } from "./context-F8WLXzPv.mjs";
 export { MODELS, app, backendModelId, buildContext, delegate, evaluateAgyPreflight, evaluateCodexPreflight, formatImageGenModelError, formatUnknownModelError, getDriver, listModelHelpLines, listRuns, nonEmptyPrompt, positiveIntSeconds, preflightCodex, preflightModel, readRunLogs, renderPreflightRefusal, resolveModel, runCli, startRun, supportsImageGen };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aibridge/cli",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "description": "CLI that bridges tasks to AI CLIs on your machine (plan / implement / review / subagent / image-gen)",
   "type": "module",
   "license": "MIT",
@@ -35,11 +35,11 @@
   ],
   "dependencies": {
     "@stricli/core": "1.3.0",
-    "@aibridge/proc": "0.4.0",
-    "@aibridge/driver-grok": "0.4.0",
-    "@aibridge/driver-agy": "0.4.0",
-    "@aibridge/driver-codex": "0.4.0",
-    "@aibridge/driver-claude": "0.4.0"
+    "@aibridge/proc": "0.5.0",
+    "@aibridge/driver-codex": "0.5.0",
+    "@aibridge/driver-claude": "0.5.0",
+    "@aibridge/driver-agy": "0.5.0",
+    "@aibridge/driver-grok": "0.5.0"
   },
   "devDependencies": {
     "tsdown": "0.22.14"

package/src/commands/subagent/command.ts

@@ -8,7 +8,7 @@ const fullDescription = [
   '',
   'Available models (canonical slug):',
   ...listModelHelpLines(),
-  'Recommended first choice: xai-grok/grok-4.5. Whichever seat runs on the same provider as the',
+  'Recommended first choice: xai-grok/grok-4.6. Whichever seat runs on the same provider as the',
   'agent you orchestrate from is your last resort — it spends the pool you are already burning.',
 ].join('\n');
 

package/src/delegate.test.ts

@@ -64,7 +64,7 @@ function createRecordingRunLog() {
 }
 
 describe('delegate stub-driver tests', () => {
-  const model = resolveModel('xai-grok/grok-4.5');
+  const model = resolveModel('xai-grok/grok-4.6');
   if (!model) throw new Error('model resolution failed');
 
   it('prepends preamble when tools: true, passes untouched when tools: false', async () => {

package/src/flagMapping.test.ts

@@ -35,7 +35,7 @@ describe('flag mapping & defaults lock', () => {
   it('plan command maps defaults correctly', async () => {
     mockPlanImpl.mockReset();
     const ctx = fakeCtx();
-    await runCli(ctx, ['plan', '--model', 'xai-grok/grok-4.5', '--out', 'plan.md', 'do something']);
+    await runCli(ctx, ['plan', '--model', 'xai-grok/grok-4.6', '--out', 'plan.md', 'do something']);
     expect(mockPlanImpl).toHaveBeenCalledTimes(1);
     const [call] = mockPlanImpl.mock.calls;
     expect(call).toBeDefined();
@@ -43,7 +43,7 @@ describe('flag mapping & defaults lock', () => {
     const [flags, prompt] = call;
     expect(prompt).toBe('do something');
     expect(flags).toEqual({
-      model: 'xai-grok/grok-4.5',
+      model: 'xai-grok/grok-4.6',
       out: 'plan.md',
       preflight: true,
     });
@@ -55,7 +55,7 @@ describe('flag mapping & defaults lock', () => {
     await runCli(ctx, [
       'plan',
       '--model',
-      'xai-grok/grok-4.5',
+      'xai-grok/grok-4.6',
       '--out',
       'plan.md',
       '--no-preflight',
@@ -70,7 +70,7 @@ describe('flag mapping & defaults lock', () => {
     const [flags, prompt] = call;
     expect(prompt).toBe('task');
     expect(flags).toEqual({
-      model: 'xai-grok/grok-4.5',
+      model: 'xai-grok/grok-4.6',
       out: 'plan.md',
       preflight: false,
       timeout: 120,
@@ -80,7 +80,7 @@ describe('flag mapping & defaults lock', () => {
   it('subagent command maps defaults correctly', async () => {
     mockSubagentImpl.mockReset();
     const ctx = fakeCtx();
-    await runCli(ctx, ['subagent', '--model', 'xai-grok/grok-4.5', 'hello agent']);
+    await runCli(ctx, ['subagent', '--model', 'xai-grok/grok-4.6', 'hello agent']);
     expect(mockSubagentImpl).toHaveBeenCalledTimes(1);
     const [call] = mockSubagentImpl.mock.calls;
     expect(call).toBeDefined();
@@ -88,7 +88,7 @@ describe('flag mapping & defaults lock', () => {
     const [flags, prompt] = call;
     expect(prompt).toBe('hello agent');
     expect(flags).toEqual({
-      model: 'xai-grok/grok-4.5',
+      model: 'xai-grok/grok-4.6',
       tools: true,
       preflight: true,
       json: false,
@@ -101,7 +101,7 @@ describe('flag mapping & defaults lock', () => {
     await runCli(ctx, [
       'subagent',
       '--model',
-      'xai-grok/grok-4.5',
+      'xai-grok/grok-4.6',
       '--no-tools',
       '--no-preflight',
       'hello agent',
@@ -113,7 +113,7 @@ describe('flag mapping & defaults lock', () => {
     const [flags, prompt] = call;
     expect(prompt).toBe('hello agent');
     expect(flags).toEqual({
-      model: 'xai-grok/grok-4.5',
+      model: 'xai-grok/grok-4.6',
       tools: false,
       preflight: false,
       json: false,

package/src/models.test.ts

@@ -10,14 +10,14 @@ import {
 
 describe('models registry', () => {
   it('resolves canonical slugs', () => {
-    const grok = resolveModel('xai-grok/grok-4.5');
+    const grok = resolveModel('xai-grok/grok-4.6');
     expect(grok).toBeDefined();
-    expect(grok?.spec.slug).toBe('xai-grok/grok-4.5');
+    expect(grok?.spec.slug).toBe('xai-grok/grok-4.6');
     expect(grok?.effort).toBeUndefined();
 
-    const gemini = resolveModel('google-antigravity/gemini-3.6-flash');
+    const gemini = resolveModel('google-antigravity/gemini-3.7-flash');
     expect(gemini).toBeDefined();
-    expect(gemini?.spec.slug).toBe('google-antigravity/gemini-3.6-flash');
+    expect(gemini?.spec.slug).toBe('google-antigravity/gemini-3.7-flash');
     expect(gemini?.effort).toBe('high');
   });
 
@@ -38,8 +38,8 @@ describe('models registry', () => {
   });
 
   it('resolves effort suffixes', () => {
-    const grokMedium = resolveModel('xai-grok/grok-4.5-medium');
-    expect(grokMedium?.spec.slug).toBe('xai-grok/grok-4.5');
+    const grokMedium = resolveModel('xai-grok/grok-4.6-medium');
+    expect(grokMedium?.spec.slug).toBe('xai-grok/grok-4.6');
     expect(grokMedium?.effort).toBe('medium');
 
     const sonnetMax = resolveModel('anthropic-claude/sonnet-5-max');
@@ -48,7 +48,7 @@ describe('models registry', () => {
   });
 
   it('rejects unsupported efforts', () => {
-    const grokXhigh = resolveModel('xai-grok/grok-4.5-xhigh');
+    const grokXhigh = resolveModel('xai-grok/grok-4.6-xhigh');
     expect(grokXhigh).toBeUndefined();
 
     const gptOssHigh = resolveModel('google-antigravity/gpt-oss-120b-medium-high');
@@ -69,27 +69,27 @@ describe('models registry', () => {
   });
 
   it('computes backendModelId correctly for agy vs others', () => {
-    const gemini = resolveModel('google-antigravity/gemini-3.6-flash');
+    const gemini = resolveModel('google-antigravity/gemini-3.7-flash');
     if (!gemini) throw new Error('gemini resolution failed');
-    expect(backendModelId(gemini)).toBe('gemini-3.6-flash-high');
+    expect(backendModelId(gemini)).toBe('gemini-3.7-flash-high');
 
-    const geminiLow = resolveModel('google-antigravity/gemini-3.6-flash-low');
+    const geminiLow = resolveModel('google-antigravity/gemini-3.7-flash-low');
     if (!geminiLow) throw new Error('geminiLow resolution failed');
-    expect(backendModelId(geminiLow)).toBe('gemini-3.6-flash-low');
+    expect(backendModelId(geminiLow)).toBe('gemini-3.7-flash-low');
 
-    const grok = resolveModel('xai-grok/grok-4.5');
+    const grok = resolveModel('xai-grok/grok-4.6');
     if (!grok) throw new Error('grok resolution failed');
-    expect(backendModelId(grok)).toBe('grok-4.5');
+    expect(backendModelId(grok)).toBe('grok-4.6');
 
     const sonnet = resolveModel('anthropic-claude/sonnet-5');
     if (!sonnet) throw new Error('sonnet resolution failed');
     expect(backendModelId(sonnet)).toBe('claude-sonnet-5');
   });
 
-  it('marks codex, grok, and gemini-3.6-flash seats as image-gen capable', () => {
+  it('marks codex, grok, and gemini-3.7-flash seats as image-gen capable', () => {
     const codex = resolveModel('openai-codex/gpt-5.6-sol');
-    const grok = resolveModel('xai-grok/grok-4.5');
-    const gemini = resolveModel('google-antigravity/gemini-3.6-flash');
+    const grok = resolveModel('xai-grok/grok-4.6');
+    const gemini = resolveModel('google-antigravity/gemini-3.7-flash');
     const claudeSonnet = resolveModel('anthropic-claude/sonnet-5');
     if (!codex || !grok || !gemini || !claudeSonnet) throw new Error('resolution failed');
     expect(supportsImageGen(codex)).toBe(true);
@@ -100,9 +100,9 @@ describe('models registry', () => {
 
   it('lists only image-capable seats when imageOnly', () => {
     const lines = listModelHelpLines({ imageOnly: true }).join('\n');
-    expect(lines).toContain('xai-grok/grok-4.5');
+    expect(lines).toContain('xai-grok/grok-4.6');
     expect(lines).toContain('openai-codex/gpt-5.6-sol');
-    expect(lines).toContain('google-antigravity/gemini-3.6-flash');
+    expect(lines).toContain('google-antigravity/gemini-3.7-flash');
     expect(lines).not.toContain('anthropic-claude/sonnet-5');
   });
 
@@ -112,9 +112,9 @@ describe('models registry', () => {
     const err = formatImageGenModelError('anthropic-claude/sonnet-5', claudeSonnet);
     expect(err).toContain('cannot generate images');
     expect(err).toContain('backend "claude"');
-    expect(err).toContain('xai-grok/grok-4.5');
+    expect(err).toContain('xai-grok/grok-4.6');
     expect(err).toContain('openai-codex/gpt-5.6-sol');
-    expect(err).toContain('google-antigravity/gemini-3.6-flash');
+    expect(err).toContain('google-antigravity/gemini-3.7-flash');
     const seatsSection = err.slice(err.indexOf('Image-gen seats'));
     expect(seatsSection).not.toContain('anthropic-claude/sonnet-5');
   });

package/src/models.ts

@@ -26,6 +26,15 @@ export interface ResolvedModel {
 }
 
 export const MODELS: Record<string, ModelSpec> = {
+  'xai-grok/grok-4.6': {
+    slug: 'xai-grok/grok-4.6',
+    backend: 'grok',
+    backendModel: 'grok-4.6',
+    efforts: ['low', 'medium', 'high'],
+    brief: 'xAI Grok 4.6 via grok CLI — own xAI login; ~30 req/min, ~1k msgs/day, single-flight',
+  },
+  // Both grok tiers stay registered: 4.6 and 4.5 differ in character, not just
+  // recency, so this is two seats of one class rather than a superseded pin.
   'xai-grok/grok-4.5': {
     slug: 'xai-grok/grok-4.5',
     backend: 'grok',
@@ -33,6 +42,17 @@ export const MODELS: Record<string, ModelSpec> = {
     efforts: ['low', 'medium', 'high'],
     brief: 'xAI Grok 4.5 via grok CLI — own xAI login; ~30 req/min, ~1k msgs/day, single-flight',
   },
+  'google-antigravity/gemini-3.7-flash': {
+    slug: 'google-antigravity/gemini-3.7-flash',
+    backend: 'agy',
+    backendModel: 'gemini-3.7-flash',
+    efforts: ['low', 'medium', 'high'],
+    defaultEffort: 'high',
+    brief:
+      'Google Gemini 3.7 Flash via agy — own Antigravity login; quota shared across all Gemini tiers (not yet itemised per-model, so no exhaustion preflight)',
+  },
+  // 3.6 stays registered while agy's quota endpoint still stops at 3.6: it is
+  // the newest flash tier the exhaustion preflight can actually guard.
   'google-antigravity/gemini-3.6-flash': {
     slug: 'google-antigravity/gemini-3.6-flash',
     backend: 'agy',

package/src/quotaPreflight.test.ts

@@ -7,6 +7,7 @@ import {
   evaluateAgyPreflight,
   evaluateCodexPreflight,
   evaluateGrokPreflight,
+  renderPreflightRefusal,
 } from './quotaPreflight.ts';
 
 test('evaluateAgyPreflight: exhausted model returns ok:false with resetTime', () => {
@@ -28,6 +29,7 @@ test('evaluateAgyPreflight: exhausted model returns ok:false with resetTime', ()
 
   assert.deepEqual(result, {
     ok: false,
+    kind: 'quota',
     message: 'agy model "Gemini 3.5 Flash (High)" is quota-exhausted',
     resetAt: '2024-01-02T12:00:00Z',
   });
@@ -127,6 +129,7 @@ test('evaluateCodexPreflight: limitReached returns ok:false', () => {
 
   assert.deepEqual(result, {
     ok: false,
+    kind: 'quota',
     message: 'codex quota limit reached',
     resetAt: '2024-01-01T17:00:00Z',
   });
@@ -199,6 +202,7 @@ test('evaluateGrokPreflight: usedPercent at 100 refuses with resetAt', () => {
 
   assert.deepEqual(result, {
     ok: false,
+    kind: 'quota',
     message: 'grok credit quota exhausted',
     resetAt: '2024-01-08T00:00:00Z',
   });
@@ -221,3 +225,27 @@ test('evaluateGrokPreflight: healthy (17% used) returns ok:true', () => {
 
   assert.deepEqual(result, { ok: true });
 });
+
+test('renderPreflightRefusal: auth kind uses unauthenticated wording', () => {
+  const msg = renderPreflightRefusal('plan', {
+    kind: 'auth',
+    message: 'grok session expired (401) — run `grok login`, then retry',
+    resetAt: undefined,
+  });
+  assert.strictEqual(
+    msg,
+    'aibridge plan: refusing — grok session expired (401) — run `grok login`, then retry. Running with --no-preflight would only send the delegate in unauthenticated. Or use a different --model.',
+  );
+});
+
+test('renderPreflightRefusal: quota kind keeps override wording', () => {
+  const msg = renderPreflightRefusal('subagent', {
+    kind: 'quota',
+    message: 'grok credit quota exhausted',
+    resetAt: undefined,
+  });
+  assert.strictEqual(
+    msg,
+    'aibridge subagent: refusing — grok credit quota exhausted. Use --no-preflight to override, or a claude-backend fallback (subagent --model sonnet|opus — bills the Claude subscription).',
+  );
+});

package/src/quotaPreflight.ts

@@ -1,11 +1,17 @@
 import { type AgyQuotaSnapshot, fetchAgyQuota, findModelQuota } from '@aibridge/driver-agy';
 import { type CodexQuotaSnapshot, fetchCodexQuota } from '@aibridge/driver-codex';
 import { fetchGrokQuota, type GrokQuotaSnapshot } from '@aibridge/driver-grok';
+import { isAuthExpired } from '@aibridge/proc';
 import { backendModelId, type ResolvedModel } from './models.ts';
 
 export type PreflightVerdict =
   | { readonly ok: true; readonly warning?: string }
-  | { readonly ok: false; readonly message: string; readonly resetAt: string | undefined };
+  | {
+      readonly ok: false;
+      readonly kind: 'auth' | 'quota';
+      readonly message: string;
+      readonly resetAt: string | undefined;
+    };
 
 export function evaluateAgyPreflight(
   snapshot: AgyQuotaSnapshot,
@@ -32,7 +38,12 @@ export function evaluateAgyPreflight(
         }
       }
     }
-    return { ok: false, message: `agy model "${backendModel}" is quota-exhausted`, resetAt };
+    return {
+      ok: false,
+      kind: 'quota',
+      message: `agy model "${backendModel}" is quota-exhausted`,
+      resetAt,
+    };
   }
 
   return { ok: true };
@@ -41,12 +52,17 @@ export function evaluateAgyPreflight(
 export function evaluateCodexPreflight(snapshot: CodexQuotaSnapshot): PreflightVerdict {
   if (snapshot.limitReached) {
     const resetAt = snapshot.windows.find(w => w.resetAt)?.resetAt;
-    return { ok: false, message: 'codex quota limit reached', resetAt };
+    return { ok: false, kind: 'quota', message: 'codex quota limit reached', resetAt };
   }
 
   const exhaustedWindow = snapshot.windows.find(w => w.usedPercent >= 100);
   if (exhaustedWindow) {
-    return { ok: false, message: 'codex quota limit reached', resetAt: exhaustedWindow.resetAt };
+    return {
+      ok: false,
+      kind: 'quota',
+      message: 'codex quota limit reached',
+      resetAt: exhaustedWindow.resetAt,
+    };
   }
 
   return { ok: true };
@@ -56,6 +72,7 @@ export function evaluateGrokPreflight(snapshot: GrokQuotaSnapshot): PreflightVer
   if (snapshot.usedPercent !== undefined && snapshot.usedPercent >= 100) {
     return {
       ok: false,
+      kind: 'quota',
       message: 'grok credit quota exhausted',
       resetAt: snapshot.periodEnd,
     };
@@ -81,10 +98,9 @@ export async function preflightModel(resolved: ResolvedModel): Promise<Preflight
     const snapshot = await fetchAgyQuota();
     return evaluateAgyPreflight(snapshot, backendModelId(resolved));
   } catch (err) {
-    return {
-      ok: true,
-      warning: `quota preflight failed (${(err as Error).message}); proceeding`,
-    };
+    if (isAuthExpired(err))
+      return { ok: false, kind: 'auth', message: err.message, resetAt: undefined };
+    return { ok: true, warning: `quota preflight failed (${(err as Error).message}); proceeding` };
   }
 }
 
@@ -93,10 +109,9 @@ export async function preflightCodex(): Promise<PreflightVerdict> {
     const snapshot = await fetchCodexQuota();
     return evaluateCodexPreflight(snapshot);
   } catch (err) {
-    return {
-      ok: true,
-      warning: `quota preflight failed (${(err as Error).message}); proceeding`,
-    };
+    if (isAuthExpired(err))
+      return { ok: false, kind: 'auth', message: err.message, resetAt: undefined };
+    return { ok: true, warning: `quota preflight failed (${(err as Error).message}); proceeding` };
   }
 }
 
@@ -105,10 +120,9 @@ export async function preflightGrok(): Promise<PreflightVerdict> {
     const snapshot = await fetchGrokQuota();
     return evaluateGrokPreflight(snapshot);
   } catch (err) {
-    return {
-      ok: true,
-      warning: `quota preflight failed (${(err as Error).message}); proceeding`,
-    };
+    if (isAuthExpired(err))
+      return { ok: false, kind: 'auth', message: err.message, resetAt: undefined };
+    return { ok: true, warning: `quota preflight failed (${(err as Error).message}); proceeding` };
   }
 }
 
@@ -124,8 +138,11 @@ function formatReset(resetTime: string | undefined): string {
 
 export function renderPreflightRefusal(
   cmd: string,
-  verdict: { message: string; resetAt: string | undefined },
+  verdict: { kind: 'auth' | 'quota'; message: string; resetAt: string | undefined },
 ): string {
+  if (verdict.kind === 'auth') {
+    return `aibridge ${cmd}: refusing — ${verdict.message}. Running with --no-preflight would only send the delegate in unauthenticated. Or use a different --model.`;
+  }
   const resetClause = verdict.resetAt ? ` Resets ${formatReset(verdict.resetAt)}.` : '';
   return `aibridge ${cmd}: refusing — ${verdict.message}.${resetClause} Use --no-preflight to override, or a claude-backend fallback (subagent --model sonnet|opus — bills the Claude subscription).`;
 }