chore: release - merge dev into main

src/backends/catalog.ts

@@ -35,7 +35,7 @@ export const CLAUDE_CODE_ENGINE_DEFINITION: AgentEngineDefinition = {
 	],
 	modelSelection: {
 		type: 'select',
-		defaultValueLabel: 'Default (Sonnet 4.5)',
+		defaultValueLabel: 'Default (Sonnet 5)',
 		options: CLAUDE_CODE_MODELS,
 	},
 	logLabel: 'Claude Code Log',

src/backends/claude-code/messageProcessing.ts

@@ -228,8 +228,11 @@ export function countToolCalls(assistantMsg: SDKAssistantMessage): number {
 /**
  * Convert a raw Anthropic model ID (e.g. 'claude-sonnet-4-5-20250929') to the
  * pricing key format used by calculateCost() (e.g. 'anthropic:claude-sonnet-4-5').
+ *
+ * Exported so the pricing-coverage drift-guard test can assert every dropdown model ID
+ * maps to a MODEL_PRICING row using the exact runtime transform (no regex re-implementation).
  */
-function toPricingKey(model: string): string {
+export function toPricingKey(model: string): string {
 	return `anthropic:${model}`.replace(/-\d{8}$/, '');
 }
 

src/backends/claude-code/models.ts

@@ -1,11 +1,16 @@
 export const CLAUDE_CODE_MODELS = [
 	{ value: 'claude-fable-5', label: 'Claude Fable 5' },
+	// Opus 5 / Sonnet 5 default to 1M context, so — like claude-fable-5 — they intentionally
+	// have no separate `[1m]` variant. (The `[1m]` suffix is a deployment identifier reserved
+	// for older generations whose default context is smaller than 1M.)
+	{ value: 'claude-opus-5', label: 'Claude Opus 5' },
 	{ value: 'claude-opus-4-8', label: 'Claude Opus 4.8' },
 	{ value: 'claude-opus-4-8[1m]', label: 'Claude Opus 4.8 (1M context)' },
 	{ value: 'claude-opus-4-7', label: 'Claude Opus 4.7' },
 	{ value: 'claude-opus-4-7[1m]', label: 'Claude Opus 4.7 (1M context)' },
 	{ value: 'claude-opus-4-6', label: 'Claude Opus 4.6' },
 	{ value: 'claude-opus-4-6[1m]', label: 'Claude Opus 4.6 (1M context)' },
+	{ value: 'claude-sonnet-5', label: 'Claude Sonnet 5' },
 	{ value: 'claude-sonnet-4-6', label: 'Claude Sonnet 4.6' },
 	{ value: 'claude-sonnet-4-6[1m]', label: 'Claude Sonnet 4.6 (1M context)' },
 	{ value: 'claude-sonnet-4-5-20250929', label: 'Claude Sonnet 4.5' },
@@ -14,4 +19,4 @@ export const CLAUDE_CODE_MODELS = [
 
 export const CLAUDE_CODE_MODEL_IDS: string[] = CLAUDE_CODE_MODELS.map((m) => m.value);
 
-export const DEFAULT_CLAUDE_CODE_MODEL = 'claude-sonnet-4-5-20250929';
+export const DEFAULT_CLAUDE_CODE_MODEL = 'claude-sonnet-5';

src/config/rateLimits.ts

@@ -26,6 +26,24 @@ export const MODEL_RATE_LIMITS: ModelRateLimits = {
 		safetyMargin: 0.85,
 	},
 
+	// Claude Opus 5 (Tier 1: 50 RPM, 10K TPM — Opus is throttle-sensitive). Needs its own row:
+	// Opus 5 draws on a rate-limit pool separate from the combined Opus 4.x pool, and the
+	// prefix matching in getRateLimitForModel would not reach an Opus 4.x row anyway. These
+	// bind on the LLMist path (getRateLimitForModel); the claude-code SDK path self-throttles.
+	'anthropic:claude-opus-5': {
+		requestsPerMinute: 50,
+		tokensPerMinute: 10_000,
+		safetyMargin: 0.85,
+	},
+
+	// Claude Sonnet 5 (Tier 1: 50 RPM, 40K TPM). Binds on the LLMist path only; the
+	// claude-code SDK path self-throttles.
+	'anthropic:claude-sonnet-5': {
+		requestsPerMinute: 50,
+		tokensPerMinute: 40_000,
+		safetyMargin: 0.9,
+	},
+
 	// Claude Opus 4.8 (Tier 1: 50 RPM, 10K TPM — Opus is throttle-sensitive)
 	'anthropic:claude-opus-4-8': {
 		requestsPerMinute: 50,

src/integrations/README.md

@@ -276,6 +276,14 @@ The fix matches on the **locale-invariant JIRA status ID** on both ends, with na
 - **Wizard** — the status-mapping select now persists the status **ID** (`{ id: s.id, name: s.name }`) while still displaying the name. `normalizeJiraStatusMappingsToIds` auto-upgrades legacy name-valued mappings → IDs in the `SET_JIRA_PROJECT_DETAILS` reducer when project details load, so re-saving any project backfills IDs. Values already-ID or unrecognized (custom) are left untouched.
 - **JQL** — `listWorkItems` quotes the status value; JIRA resolves a quoted numeric value against status IDs, so ID-based config values remain valid with no behavior change.
 
+#### JIRA issue-type mapping is read from `issueTypes.task` (MNG-1769)
+
+Same "wizard writes X → runtime reads X" failure shape as the locale-fragile status bug above ([MNG-1768](https://linear.app/mongrel/issue/MNG-1768/jira-status-matching-is-locale-fragile-status-moves-silently-no-op)). The JIRA wizard's `IssueTypeMappingStep` (`web/src/components/projects/pm-providers/jira/issue-type-step.tsx`) persists the operator's Task mapping under `jira.issueTypes.task`, but `JiraPMProvider.createWorkItem` used to read `issueTypes.default` — a key nothing ever wrote — so the optional chain always yielded `undefined` and **every** JIRA issue was hardcoded to type `"Task"`, silently ignoring the operator's mapping.
+
+- **Runtime reads `issueTypes.task`.** `createWorkItem` reads `this.config.issueTypes?.task ?? 'Task'`. The `'Task'` fallback is retained for configs that never set a mapping (backward compatible). The legacy `issueTypes.default` key is intentionally **not** read — honoring it would resurrect the bug — and a regression test in `tests/unit/pm/jira/adapter.test.ts` proves `default` is no longer honored.
+- **Actionable failure.** When `jiraClient.createIssue` fails (commonly a JIRA 400 when the mapped/fallback type does not exist on the project), the adapter best-effort calls `jiraClient.getIssueTypesForProject(projectKey)` and re-throws an error naming the attempted type and the project's discovered non-subtask issue types. The diagnostic fetch is guarded so a discovery failure re-throws the original creation error unchanged.
+- **`subtask` is intentionally not consumed.** There is no subtask-creation path, so the wizard's subtask row was removed rather than persist config nothing reads. A previously-saved `issueTypes.subtask` value is harmless — it simply stays in config, unread.
+
 ### Wizard path — metadata-driven, shared between providers
 
 The PM wizards consume the workflow status definition list through a single tRPC query (`trpc.workflowStatuses.list`) and render mapping rows for every key — built-in and custom alike. The provider's `useProviderHooks` resolves the list and forwards it as `workflowStatuses` on the hook return; the shared `StatusMappingStep` renders rows in the returned order. Reference implementations:

src/pm/jira/adapter.ts

@@ -46,6 +46,16 @@ import { adfToPlainText, extractAdfMediaNodes, markdownToAdf } from './adf.js';
  * enforces invariance on object-property types. Internally the adapter
  * uses `config.containerId` as a JIRA project key — the project-scoped
  * entry point.
+ *
+ * MNG-1769: `createWorkItem` reads the operator's Task mapping from
+ * `issueTypes.task` — the exact key the wizard's `IssueTypeMappingStep`
+ * writes (`web/src/components/projects/pm-providers/jira/issue-type-step.tsx`).
+ * The `'Task'` string is the last-resort fallback for configs that never set
+ * a mapping. The legacy `issueTypes.default` key is intentionally NOT read —
+ * nothing ever wrote it, so honoring it would resurrect the bug where every
+ * JIRA issue was hardcoded to type `"Task"`. `subtask` is intentionally not
+ * consumed: there is no subtask-creation path. This is the "wizard writes X →
+ * runtime reads X" contract, mirroring MNG-1768's parity guard.
  */
 
 interface JiraConfig {
@@ -194,14 +204,23 @@ export class JiraPMProvider implements PMProvider {
 	}
 
 	async createWorkItem(config: CreateWorkItemConfig): Promise<WorkItem> {
-		const issueType = this.config.issueTypes?.default ?? 'Task';
-		const result = await jiraClient.createIssue({
-			project: { key: config.containerId || this.config.projectKey },
-			summary: config.title,
-			description: config.description ? markdownToAdf(config.description) : undefined,
-			issuetype: { name: issueType },
-			...(config.labels?.length ? { labels: config.labels } : {}),
-		});
+		// MNG-1769: read the key the wizard actually writes (`issueTypes.task`),
+		// not the never-written legacy `issueTypes.default`. `'Task'` remains the
+		// last-resort fallback for configs that never set a mapping.
+		const issueType = this.config.issueTypes?.task ?? 'Task';
+		const projectKey = config.containerId || this.config.projectKey;
+		let result: Awaited<ReturnType<typeof jiraClient.createIssue>>;
+		try {
+			result = await jiraClient.createIssue({
+				project: { key: projectKey },
+				summary: config.title,
+				description: config.description ? markdownToAdf(config.description) : undefined,
+				issuetype: { name: issueType },
+				...(config.labels?.length ? { labels: config.labels } : {}),
+			});
+		} catch (err) {
+			throw await this.enrichCreateIssueError(err, projectKey, issueType);
+		}
 		const key = result.key ?? '';
 
 		// Transition to backlog status if configured
@@ -230,6 +249,45 @@ export class JiraPMProvider implements PMProvider {
 		};
 	}
 
+	/**
+	 * MNG-1769: when `jiraClient.createIssue` fails (commonly a JIRA 400 because
+	 * the configured/fallback issue type does not exist on the project), best-effort
+	 * fetch the project's discovered issue types and re-throw an Error naming the
+	 * attempted type and the available non-subtask types. This turns an opaque JIRA
+	 * 400 into an actionable message pointing the operator at the wizard mapping.
+	 *
+	 * The diagnostic fetch is guarded in its own try/catch so a discovery failure
+	 * never masks the original creation error — if discovery also fails, the
+	 * original error surfaces unchanged.
+	 */
+	private async enrichCreateIssueError(
+		originalError: unknown,
+		projectKey: string,
+		attemptedType: string,
+	): Promise<unknown> {
+		try {
+			const types = await jiraClient.getIssueTypesForProject(projectKey);
+			const available = types
+				.filter((t) => !t.subtask)
+				.map((t) => t.name)
+				.filter((name) => name.length > 0);
+			return new Error(
+				`Failed to create JIRA issue in project "${projectKey}" with issue type "${attemptedType}". ` +
+					`Available issue types for this project: ${
+						available.length > 0 ? available.join(', ') : '(none discovered)'
+					}. ` +
+					`Map the Task role to one of these in the JIRA wizard's issue-type step. ` +
+					`Original error: ${String(originalError)}`,
+			);
+		} catch (discoveryErr) {
+			logger.warn('[JIRA] Could not fetch issue types while enriching createIssue error', {
+				projectKey,
+				error: String(discoveryErr),
+			});
+			return originalError;
+		}
+	}
+
 	async listWorkItems(
 		containerId: ContainerId | undefined,
 		filter?: ListWorkItemsFilter,

src/utils/llmMetrics.ts

@@ -3,27 +3,54 @@
  * Provides cost calculation.
  */
 import type { TokenUsage } from 'llmist';
+import { captureException } from '../sentry.js';
+import { logger } from './logging.js';
+
+/**
+ * Models we've already warned about this process, so a missing-pricing row logs/captures
+ * once per unique model instead of on every LLM call/turn. Workers are ephemeral (one job
+ * per container), so process-level dedup is the right granularity.
+ */
+const warnedMissingPricing = new Set<string>();
 
 /**
  * Model pricing per 1M tokens (in USD).
- * Prices as of January 2026.
+ * Prices as of August 2026.
  */
-const MODEL_PRICING: Record<string, { input: number; output: number; cachedInput?: number }> = {
+export const MODEL_PRICING: Record<
+	string,
+	{ input: number; output: number; cachedInput?: number }
+> = {
 	// Anthropic Claude Fable 5 — 1M context by default (max = default), priced at 2× Opus.
 	// Key matches toPricingKey('claude-fable-5') = 'anthropic:claude-fable-5' (no trailing
 	// date to strip). cachedInput follows the 0.1× convention used by every Anthropic row.
 	'anthropic:claude-fable-5': { input: 10.0, output: 50.0, cachedInput: 1.0 },
 
+	// Anthropic Claude 5 family
+	// Opus 5: mirrors Opus 4.8 pricing; cachedInput follows the 0.1× Anthropic convention.
+	'anthropic:claude-opus-5': { input: 5.0, output: 25.0, cachedInput: 0.5 },
+	// Sonnet 5: seeded at standard post-intro rates. Intro pricing is $2.00/$10.00 per MTok
+	// through 2026-08-31; seeding at standard rates over-reports during the intro window,
+	// which is the safe direction for budgets (never under-reports). Revisit before that
+	// date if a short-lived intro-rate edit is wanted.
+	'anthropic:claude-sonnet-5': { input: 3.0, output: 15.0, cachedInput: 0.3 },
+
 	// Anthropic Claude 4 family
 	'anthropic:claude-opus-4-8': { input: 5.0, output: 25.0, cachedInput: 0.5 },
 	'anthropic:claude-opus-4-8[1m]': { input: 5.0, output: 25.0, cachedInput: 0.5 },
 	'anthropic:claude-opus-4-7': { input: 5.0, output: 25.0, cachedInput: 0.5 },
 	'anthropic:claude-opus-4-7[1m]': { input: 5.0, output: 25.0, cachedInput: 0.5 },
+	// Bare claude-opus-4-6 backfill (was only present as the [1m] variant, so the bare
+	// dropdown ID ran unpriced at $0 — a silent budget bypass). Mirrors the [1m] row.
+	'anthropic:claude-opus-4-6': { input: 5.0, output: 25.0, cachedInput: 0.5 },
 	'anthropic:claude-opus-4-6[1m]': { input: 5.0, output: 25.0, cachedInput: 0.5 },
 	'anthropic:claude-sonnet-4-6': { input: 3.0, output: 15.0, cachedInput: 0.3 },
 	'anthropic:claude-sonnet-4-6[1m]': { input: 3.0, output: 15.0, cachedInput: 0.3 },
 	'anthropic:claude-sonnet-4-5': { input: 3.0, output: 15.0, cachedInput: 0.3 },
 	'anthropic:claude-opus-4-5': { input: 15.0, output: 75.0, cachedInput: 1.5 },
+	// Bare claude-haiku-4-5 backfill: claude-haiku-4-5-20251001 is in the dropdown but
+	// only claude-haiku-3-5 was priced, so Haiku 4.5 ran unpriced at $0.
+	'anthropic:claude-haiku-4-5': { input: 1.0, output: 5.0, cachedInput: 0.1 },
 	'anthropic:claude-haiku-3-5': { input: 0.8, output: 4.0, cachedInput: 0.08 },
 
 	// Google Gemini
@@ -74,7 +101,26 @@ const MODEL_PRICING: Record<string, { input: number; output: number; cachedInput
  */
 export function calculateCost(model: string, usage: TokenUsage): number {
 	const pricing = MODEL_PRICING[model];
-	if (!pricing) return 0;
+	if (!pricing) {
+		// A missing pricing row makes calculateCost return 0, which silently disables
+		// workItemBudget enforcement for that model (checkBudgetExceeded never trips on a
+		// $0 spend). Make the miss loud — but non-fatal: calculateCost is a hot-path pure
+		// utility called across all three engines, so throwing would crash runs for any
+		// model not yet in MODEL_PRICING. Loud-observe + the drift-guard test is the safer
+		// combination. Dedup so it fires once per unique model per process, not per turn.
+		if (!warnedMissingPricing.has(model)) {
+			warnedMissingPricing.add(model);
+			logger.warn(
+				`No MODEL_PRICING row for "${model}"; cost reported as $0. This silently disables workItemBudget enforcement for this model — add a pricing row in src/utils/llmMetrics.ts.`,
+			);
+			captureException(new Error(`Missing MODEL_PRICING row for model "${model}"`), {
+				tags: { source: 'model_pricing_missing' },
+				level: 'warning',
+				extra: { model },
+			});
+		}
+		return 0;
+	}
 
 	const inputCost = (usage.inputTokens / 1_000_000) * pricing.input;
 	const outputCost = (usage.outputTokens / 1_000_000) * pricing.output;

tests/unit/backends/claude-code.test.ts

@@ -29,13 +29,15 @@ import {
 	resolveClaudeCodeExecutablePath,
 	resolveClaudeModel,
 } from '../../../src/backends/claude-code/index.js';
+import { toPricingKey } from '../../../src/backends/claude-code/messageProcessing.js';
 import {
 	CLAUDE_CODE_MODEL_IDS,
 	CLAUDE_CODE_MODELS,
 	DEFAULT_CLAUDE_CODE_MODEL,
 } from '../../../src/backends/claude-code/models.js';
 import { resolveClaudeCodeSettings } from '../../../src/backends/claude-code/settings.js';
 import type { AgentExecutionPlan, ToolManifest } from '../../../src/backends/types.js';
+import { MODEL_PRICING } from '../../../src/utils/llmMetrics.js';
 
 const mockQuery = vi.mocked(query);
 
@@ -295,7 +297,15 @@ describe('buildSystemPrompt', () => {
 
 describe('CLAUDE_CODE_MODELS constants', () => {
 	it('contains the expected models', () => {
-		expect(CLAUDE_CODE_MODELS).toHaveLength(11);
+		expect(CLAUDE_CODE_MODELS).toHaveLength(13);
+	});
+
+	it('includes Opus 5 and Sonnet 5 (1M context by default — no [1m] variant)', () => {
+		expect(CLAUDE_CODE_MODEL_IDS).toContain('claude-opus-5');
+		expect(CLAUDE_CODE_MODEL_IDS).toContain('claude-sonnet-5');
+		// Opus 5 / Sonnet 5 are 1M-context by default, so a [1m] suffix would be redundant.
+		expect(CLAUDE_CODE_MODEL_IDS).not.toContain('claude-opus-5[1m]');
+		expect(CLAUDE_CODE_MODEL_IDS).not.toContain('claude-sonnet-5[1m]');
 	});
 
 	it('includes Opus 4.8, Opus 4.7, and the 1M context variants', () => {
@@ -329,9 +339,26 @@ describe('CLAUDE_CODE_MODELS constants', () => {
 	});
 });
 
+describe('MODEL_PRICING coverage (drift guard)', () => {
+	// This single assertion would have caught the pre-existing bare claude-opus-4-6 and
+	// claude-haiku-4-5 pricing gaps, and will catch the next one. A missing MODEL_PRICING
+	// row makes calculateCost return $0, which silently disables workItemBudget enforcement.
+	it('every CLAUDE_CODE_MODEL_IDS entry maps to a MODEL_PRICING row', () => {
+		for (const id of CLAUDE_CODE_MODEL_IDS) {
+			const pricingKey = toPricingKey(id);
+			expect(
+				MODEL_PRICING[pricingKey],
+				`${id} → ${pricingKey} has no MODEL_PRICING row`,
+			).toBeDefined();
+		}
+	});
+});
+
 describe('resolveClaudeModel', () => {
 	it('passes through known Claude Code model IDs', () => {
 		expect(resolveClaudeModel('claude-fable-5')).toBe('claude-fable-5');
+		expect(resolveClaudeModel('claude-opus-5')).toBe('claude-opus-5');
+		expect(resolveClaudeModel('claude-sonnet-5')).toBe('claude-sonnet-5');
 		expect(resolveClaudeModel('claude-opus-4-8')).toBe('claude-opus-4-8');
 		expect(resolveClaudeModel('claude-opus-4-8[1m]')).toBe('claude-opus-4-8[1m]');
 		expect(resolveClaudeModel('claude-opus-4-7')).toBe('claude-opus-4-7');

tests/unit/config/rateLimits.test.ts

@@ -45,6 +45,26 @@ describe.concurrent('config/rateLimits', () => {
 			});
 		});
 
+		it('returns exact match for Claude Opus 5', () => {
+			const result = getRateLimitForModel('anthropic:claude-opus-5');
+
+			expect(result).toEqual({
+				requestsPerMinute: 50,
+				tokensPerMinute: 10_000,
+				safetyMargin: 0.85,
+			});
+		});
+
+		it('returns exact match for Claude Sonnet 5', () => {
+			const result = getRateLimitForModel('anthropic:claude-sonnet-5');
+
+			expect(result).toEqual({
+				requestsPerMinute: 50,
+				tokensPerMinute: 40_000,
+				safetyMargin: 0.9,
+			});
+		});
+
 		it('returns prefix match for models with version suffix', () => {
 			// anthropic:claude-sonnet-4-5-20250929 should match anthropic:claude-sonnet-4-5
 			const result = getRateLimitForModel('anthropic:claude-sonnet-4-5-20250929');
@@ -120,6 +140,11 @@ describe.concurrent('config/rateLimits', () => {
 			expect(MODEL_RATE_LIMITS['anthropic:claude-opus-4-5']).toBeDefined();
 		});
 
+		it('includes Claude Opus 5 and Sonnet 5 configs', () => {
+			expect(MODEL_RATE_LIMITS['anthropic:claude-opus-5']).toBeDefined();
+			expect(MODEL_RATE_LIMITS['anthropic:claude-sonnet-5']).toBeDefined();
+		});
+
 		it('includes OpenRouter models', () => {
 			expect(MODEL_RATE_LIMITS['openrouter:google/gemini-3-flash-preview']).toBeDefined();
 			expect(MODEL_RATE_LIMITS['openrouter:deepseek/deepseek-chat-v3-0324']).toBeDefined();