@bastani/atomic 0.9.13 → 0.9.14-alpha.2

package/dist/builtin/workflows/builtin/generate-and-filter.ts

@@ -6,6 +6,10 @@ import { runGenerateAndFilter } from "./generate-and-filter-runner.js";
 export default workflow({
   name: "generate-and-filter",
   description: "Generate more independent candidates than needed, deduplicate and filter them by rubric, optionally judge them, and return a parent-consumable shortlist.",
+  // The 15-minute default, stated rather than inherited: this is a per-workflow
+  // product decision, so a future change to the global default must not silently
+  // re-cadence a long autonomous run.
+  heartbeatIntervalMinutes: 15,
   inputs: {
     prompt: Type.String({ description: "Prompt for candidate generation and selection." }),
     num_candidates: Type.Integer({ minimum: 2, maximum: 20, default: 8, description: "Number of independent candidates to generate." }),

package/dist/builtin/workflows/builtin/goal-models.ts

@@ -19,9 +19,9 @@ export const orchestratorModelConfig = {
       "openai/gpt-5.5:xhigh",
       "anthropic/claude-opus-4-8:high",
       "github-copilot/claude-opus-4.8:high",
-      "xai/grok-4.5:high",
-      "zai/glm-5.2:xhigh",
-      "zai-coding-cn/glm-5.2:xhigh",
+      "xai/grok-4.6:high",
+      "zai/glm-5.3:high",
+      "zai-coding-cn/glm-5.3:high",
       "openrouter/anthropic/claude-opus-5:high",
       "openrouter/openai/gpt-5.6-sol:xhigh",
       "openrouter/anthropic/claude-fable-5:high",
@@ -29,8 +29,7 @@ export const orchestratorModelConfig = {
       "openrouter/sakana/fugu-ultra:high",
       "openrouter/openai/gpt-5.5:xhigh",
       "openrouter/anthropic/claude-opus-4-8:high",
-      "openrouter/x-ai/grok-4.5",
-      "openrouter/z-ai/glm-5.2:xhigh"
+      "openrouter/x-ai/grok-4.6",
     ],
     excludedTools: ["ask_user_question"],
 };
@@ -54,9 +53,9 @@ export const reviewerModelConfig = {
       "openai/gpt-5.5:xhigh",
       "anthropic/claude-opus-4-8:high",
       "github-copilot/claude-opus-4.8:high",
-      "xai/grok-4.5:high",
-      "zai/glm-5.2:xhigh",
-      "zai-coding-cn/glm-5.2:xhigh",
+      "xai/grok-4.6:high",
+      "zai/glm-5.3:high",
+      "zai-coding-cn/glm-5.3:high",
       "openrouter/anthropic/claude-opus-5:high",
       "openrouter/anthropic/claude-fable-5:high",
       "openrouter/openai/gpt-5.6-sol:xhigh",
@@ -64,8 +63,7 @@ export const reviewerModelConfig = {
       "openrouter/sakana/fugu-ultra:high",
       "openrouter/openai/gpt-5.5:xhigh",
       "openrouter/anthropic/claude-opus-4-8:high",
-      "openrouter/x-ai/grok-4.5",
-      "openrouter/z-ai/glm-5.2:xhigh"
+      "openrouter/x-ai/grok-4.6",
     ],
     excludedTools: ["ask_user_question"],
     schema: reviewDecisionSchema,

package/dist/builtin/workflows/builtin/goal.ts

@@ -15,6 +15,10 @@ import { DEFAULT_MAX_TURNS } from "./goal-types.js";
 export default workflow({
   name: "goal",
   description: "Goal Runner workflow with bounded sub-agent orchestration turns, immutable acceptance criteria, ledger artifacts, parallel reviewers, and reducer-gated completion. When launching follow-up goal runs from review findings, pass the ORIGINAL task text as acceptance_criteria so deltas cannot drift from the literal contract. If the task includes submitting a pull request (or MR/review), remove that final action from the objective text and set create_pr=true instead when preparing the workflow inputs.",
+  // The 15-minute default, stated rather than inherited: this is a per-workflow
+  // product decision, so a future change to the global default must not silently
+  // re-cadence a long autonomous run.
+  heartbeatIntervalMinutes: 15,
   inputs: {
     objective: Type.String({ description: "The objective or delta for this Goal Runner workflow run. Do not include PR/MR submission instructions here; strip them from the task text and request them via create_pr=true instead." }),
     acceptance_criteria: Type.Optional(Type.String({ description: "Original immutable task contract this run must remain consistent with. Defaults to objective. Orchestrators launching follow-up runs from reviewer findings should pass the ORIGINAL task text here." })),

package/dist/builtin/workflows/builtin/index.d.ts

@@ -21,20 +21,16 @@ export type OpenClaudeDesignOutputType = "prototype" | "wireframe" | "page" | "c
 export type OpenClaudeDesignWorkflowInputs = {
   readonly prompt: string;
   readonly discover_references: boolean;
-  readonly max_refinements: number;
 };
 export type OpenClaudeDesignWorkflowRunInputs = {
   readonly prompt: string;
   readonly discover_references?: boolean;
-  readonly max_refinements?: number;
 };
 export type OpenClaudeDesignWorkflowOutputs = WorkflowOutputValues & {
   readonly output_type?: string;
   readonly design_system?: string;
   readonly artifact?: string;
   readonly handoff?: string;
-  readonly approved_for_export?: boolean;
-  readonly refinements_completed?: number;
   readonly import_context?: string;
   readonly run_id?: string;
   readonly artifact_dir?: string;

package/dist/builtin/workflows/builtin/loop-until-done-runner.ts

@@ -129,10 +129,12 @@ export async function runLoopUntilDone(ctx: WorkflowRunContext<LoopInputs>) {
       context: "fresh",
       reads: [ledgerPath, iterationPath],
       schema: evaluationSchema,
-      output: evaluationPath,
-      outputMode: "file-only",
     });
     const decision = evaluationFrom(evaluator);
+    // Evaluation reports are declared workflow outputs consumed as data: the
+    // runner persists the structured decision itself so evaluation-N.json
+    // stays schema-shaped JSON rather than stage prose.
+    await writeFile(evaluationPath, `${JSON.stringify(evaluator.structured, null, 2)}\n`);
     entries.push({
       iteration,
       artifact_path: iterationPath,

package/dist/builtin/workflows/builtin/loop-until-done.ts

@@ -6,6 +6,10 @@ import { runLoopUntilDone } from "./loop-until-done-runner.js";
 export default workflow({
   name: "loop-until-done",
   description: "Repeat evidence-producing work and independent completion evaluation against a durable ledger until done or an inspectable iteration-limit failure.",
+  // The 15-minute default, stated rather than inherited: this is a per-workflow
+  // product decision, so a future change to the global default must not silently
+  // re-cadence a long autonomous run.
+  heartbeatIntervalMinutes: 15,
   inputs: {
     prompt: Type.String({ description: "Objective whose explicit completion condition controls the bounded loop." }),
     max_iterations: Type.Integer({

package/dist/builtin/workflows/builtin/open-claude-design-live-protocol.ts

@@ -0,0 +1,305 @@
+/**
+ * open-claude-design ⇄ impeccable live protocol.
+ *
+ * In a chat session the agent owns the live poll loop, and "the review is over"
+ * needs no ceremony because the user simply keeps talking. A workflow stage
+ * boundary is a commit point, and the live contract's `"timeout" -> LOOP` is an
+ * instruction to a model rather than a constraint: a stage that concludes on a
+ * poll timeout would end a review the user never finished.
+ *
+ * So the workflow drives the protocol instead. Polling and replying are durable
+ * `ctx.tool` nodes; the model is invoked only for the events that need it
+ * (`generate`, `steer`, `manual_edit_apply`, and mount failures). Termination
+ * stops being a judgment and becomes the helper's own `exit` event.
+ *
+ * `timeout` is absorbed here rather than surfaced, so an idle hour costs no
+ * graph nodes. Every poll/reply invocation must, however, exit successfully;
+ * helper failures are surfaced instead of being mistaken for another timeout.
+ */
+
+import { spawn } from "node:child_process";
+import { fileURLToPath } from "node:url";
+
+/** JSON values emitted by the live helper. */
+export type JsonValue = string | number | boolean | null | readonly JsonValue[] | { readonly [key: string]: JsonValue };
+
+/** Events the live helper delivers. Anything unknown is treated as ignorable. */
+export type LiveEventType =
+	| "generate"
+	| "steer"
+	| "accept"
+	| "discard"
+	| "prefetch"
+	| "manual_edit_apply"
+	| "variant_mounted"
+	| "variant_mount_failed"
+	| "timeout"
+	| "exit";
+
+export type LiveEvent = {
+	readonly type: LiveEventType | string;
+	readonly id?: string;
+	readonly variant?: number;
+	readonly variantId?: string;
+	readonly url?: string;
+	readonly pageUrl?: string;
+	readonly error?: string;
+	readonly file?: string;
+	readonly sourceFile?: string;
+	readonly message?: string;
+	readonly data?: JsonValue;
+	readonly screenshot?: string;
+	readonly raw: string;
+	/** Preserve additional upstream event fields without narrowing their JSON shape. */
+	readonly [key: string]: JsonValue | undefined;
+};
+
+/** Events the model must handle before the loop may reply and poll again. */
+const MODEL_EVENTS = new Set<string>(["generate", "steer", "manual_edit_apply", "variant_mount_failed"]);
+
+export function needsModel(event: LiveEvent): boolean {
+	return MODEL_EVENTS.has(event.type);
+}
+
+/** Reply status the helper expects for each model-handled event. */
+export function replyTokenFor(event: LiveEvent): string {
+	if (event.type === "steer") return "steer_done";
+	return "done";
+}
+
+const LEGACY_REPLY_STATUSES = new Set(["done", "error", "complete", "discard", "discarded"]);
+
+function isJsonObject(value: JsonValue): value is { readonly [key: string]: JsonValue } {
+	return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+
+/** Last JSON object printed by a live script, which is the event payload. */
+export function parseLiveEvent(stdout: string): LiveEvent {
+	const lines = stdout.split(/\r?\n/).filter((line) => line.trim().startsWith("{"));
+	for (const line of [...lines].reverse()) {
+		try {
+			const parsed = JSON.parse(line) as JsonValue;
+			if (!isJsonObject(parsed) || typeof parsed.type !== "string") continue;
+			// Keep the complete JSON event, not only the fields known by the old
+			// protocol. 4.1.1 adds mount diagnostics and structured reply data.
+			return { ...parsed, type: parsed.type, raw: line } as LiveEvent;
+		} catch {
+			/* not the event line */
+		}
+	}
+	// An unreadable poll is not an ending. Treat it as a timeout so the caller
+	// polls again rather than concluding a review nobody finished.
+	return { type: "timeout", raw: stdout };
+}
+
+export type LivePollDeps = {
+	readonly run?: (scriptPath: string, args: readonly string[], cwd: string, signal?: AbortSignal) => Promise<ScriptResult>;
+};
+
+export type LiveReplyInput = {
+	readonly workflowCwd: string;
+	readonly event: LiveEvent;
+	readonly status?: string;
+	/** Source/manifest path accepted by generate and mount-failure replies. */
+	readonly file?: string;
+	/** The helper's positional message argument (there is no --message flag). */
+	readonly message?: string;
+	/** Structured result accepted by manual_edit_apply replies. */
+	readonly data?: JsonValue;
+	readonly signal?: AbortSignal;
+	readonly deps?: LivePollDeps;
+};
+
+function assertSuccessfulScript(result: ScriptResult, operation: string): ScriptResult {
+	if (result.code !== 0) {
+		const details = result.stderr.length > 0 ? `: ${result.stderr}` : "";
+		throw new Error(`${operation} failed with exit code ${result.code}${details}`);
+	}
+	return result;
+}
+
+/**
+ * Poll until the helper delivers an event that matters. `timeout` events and
+ * unreadable output loop internally, so callers only ever see substantive
+ * events or `exit`.
+ */
+export async function pollLiveEvent(input: {
+	readonly workflowCwd: string;
+	readonly signal?: AbortSignal;
+	readonly deps?: LivePollDeps;
+}): Promise<LiveEvent> {
+	const script = liveScriptPath("live-poll.mjs");
+	const run = input.deps?.run ?? runNodeScript;
+	for (;;) {
+		if (input.signal?.aborted === true) return { type: "exit", raw: "aborted" };
+		const result = assertSuccessfulScript(await run(script, [], input.workflowCwd, input.signal), "live poll");
+		const event = parseLiveEvent(result.stdout);
+		if (event.type !== "timeout") return event;
+	}
+}
+
+/** Acknowledge a model-handled event so the helper unblocks the overlay. */
+export async function replyLiveEvent(input: LiveReplyInput): Promise<ScriptResult> {
+	const id = input.event.id;
+	if (typeof id !== "string" || id.length === 0 || id.startsWith("--")) {
+		throw new Error(`Cannot reply to ${input.event.type}: live event id is missing`);
+	}
+	if (LEGACY_REPLY_STATUSES.has(id)) {
+		throw new Error(`Cannot reply to live event ${id}: expected an event id before the reply status`);
+	}
+	const status = input.status ?? replyTokenFor(input.event);
+	if (status.length === 0 || status.startsWith("--")) {
+		throw new Error(`Cannot reply to live event ${id}: reply status is missing`);
+	}
+	const args = ["--reply", id, status];
+	const file = input.file ?? input.event.file;
+	const message = input.message ?? input.event.message;
+	const data = input.data ?? input.event.data;
+	if (file !== undefined) args.push("--file", file);
+	if (data !== undefined) args.push("--data", JSON.stringify(data));
+	if (message !== undefined) args.push(message);
+	const script = liveScriptPath("live-poll.mjs");
+	const run = input.deps?.run ?? runNodeScript;
+	return assertSuccessfulScript(await run(script, args, input.workflowCwd, input.signal), "live reply");
+}
+
+/**
+ * Absolute path to one of the impeccable live scripts.
+ *
+ * Always the copy that ships inside this package — `packages/workflows/skills`
+ * in the repository, `dist/builtin/workflows/skills` once bundled — which sits
+ * one level up from this module in both layouts, so `import.meta.url` finds it
+ * without the host naming a path.
+ *
+ * A project-vendored copy is deliberately NOT consulted. The workflow drives
+ * this protocol directly: it depends on `live-poll.mjs`'s CLI surface, its
+ * `--reply` tokens, and its JSON event vocabulary. The bundled scripts ship and
+ * are tested with this code, so they cannot drift from it; a forked copy in a
+ * project could, and would break the loop in ways no test here would catch.
+ */
+export function liveScriptPath(script: string): string {
+	return fileURLToPath(new URL(`../skills/impeccable/scripts/${script}`, import.meta.url));
+}
+
+export type ScriptResult = { readonly code: number; readonly stdout: string; readonly stderr: string };
+
+function runNodeScript(
+	scriptPath: string,
+	args: readonly string[],
+	cwd: string,
+	signal?: AbortSignal,
+): Promise<ScriptResult> {
+	return new Promise((resolve, reject) => {
+		const child = spawn(process.execPath, [scriptPath, ...args], {
+			cwd,
+			env: { ...process.env, GH_PAGER: "cat" },
+			stdio: ["ignore", "pipe", "pipe"],
+		});
+		let stdout = "";
+		let stderr = "";
+		let settled = false;
+		const onAbort = () => child.kill("SIGTERM");
+		signal?.addEventListener("abort", onAbort, { once: true });
+		child.stdout.on("data", (chunk: Buffer) => {
+			stdout += chunk.toString();
+		});
+		child.stderr.on("data", (chunk: Buffer) => {
+			stderr += chunk.toString();
+		});
+		child.on("error", (error) => {
+			if (settled) return;
+			settled = true;
+			signal?.removeEventListener("abort", onAbort);
+			reject(error);
+		});
+		child.on("close", (code) => {
+			if (settled) return;
+			settled = true;
+			signal?.removeEventListener("abort", onAbort);
+			resolve({ code: code ?? -1, stdout: stdout.trim(), stderr: stderr.trim() });
+		});
+	});
+}
+/** Prompt for the stage that handles one live event. */
+export function buildLiveEventPrompt(input: {
+	readonly event: LiveEvent;
+	readonly previewPath: string;
+}): string {
+	const shared = [
+		`<live_event>\n${input.event.raw}\n</live_event>`,
+		`<preview_path>${input.previewPath}</preview_path>`,
+		"<role>You are an opinionated staff design engineer inside an impeccable `live` session. The workflow owns the poll loop; handle exactly this one event and stop.</role>",
+	];
+	if (input.event.type === "generate") {
+		return [
+			...shared,
+			[
+				"<instructions>",
+				"Follow the live contract for a `generate` event:",
+				"1. Reuse `event.scaffold` when present and read the annotation screenshot when one is attached.",
+				"2. Extract the page identity FIRST, then produce three DISTINCT on-brand variants and publish them for the user to preview.",
+				"3. Apply the craft floors by construction as you write. Do not screenshot, re-render, or QA between generate and accept; the overlay preview is the verification channel.",
+				"Do not poll, do not reply, and do not exit the session: the workflow does that. Stop as soon as the variants are published.",
+				"</instructions>",
+			].join("\n"),
+		].join("\n\n");
+	}
+	if (input.event.type === "steer") {
+		return [
+			...shared,
+			[
+				"<instructions>",
+				"Follow the live contract for a `steer` event: read the message and pageUrl, then do the work — page edits, navigation help, or a short answer.",
+				"Do not poll, do not reply, and do not exit the session: the workflow does that. Stop when the steer is handled.",
+				"</instructions>",
+			].join("\n"),
+		].join("\n\n");
+	}
+	if (input.event.type === "variant_mount_failed") {
+		return [
+			...shared,
+			[
+				"<instructions>",
+				"The browser could not mount the published variant. Inspect the reported variant source/module, fix the compile or runtime failure, and stop after the repair.",
+				"Do not poll, do not reply, and do not exit the session: the workflow does that.",
+			].join("\n"),
+		].join("\n\n");
+	}
+	return [
+		...shared,
+		[
+			"<instructions>",
+			"Apply this manual edit event per the live contract, then stop. Do not poll, reply, or exit; the workflow does that.",
+			"</instructions>",
+		].join("\n"),
+	].join("\n\n");
+}
+
+/** Prompt for the stage that boots the session and hands the loop back to the workflow. */
+export function buildLiveSessionStartPrompt(input: {
+	readonly previewPath: string;
+	readonly previewFileUrl: string;
+	readonly browserBootstrapRules: string;
+	readonly round: number;
+}): string {
+	return [
+		`<preview_path>${input.previewPath}</preview_path>`,
+		`<preview_file_url>${input.previewFileUrl}</preview_file_url>`,
+		`<browser_use_guidelines>${input.browserBootstrapRules}</browser_use_guidelines>`,
+		"<role>You are an opinionated staff design engineer opening an impeccable `live` review session.</role>",
+		[
+			"<objective>",
+			`Start review session ${input.round}: boot live against the static preview and open it in the user's browser. Then STOP.`,
+			"</objective>",
+		].join("\n"),
+		[
+			"<instructions>",
+			"1. Drive `/skill:impeccable live` against the static preview: run `live.mjs` with `--target` pointed at the preview file, or the equivalent `.impeccable/live/config.json` entry, and open the URL that serves it.",
+			"2. Print the live `http://` review URL in plain text, plus the preview file URL as the manual fallback, so anyone attaching to this run can find the review.",
+			"3. Directly under the URL, print how the user ends the review, in plain text: click exit in the Impeccable overlay, close the browser tab, or say `exit live`. State that the review waits indefinitely until they do, and that ending it exports the design as it then stands, with no further round.",
+			"4. Do NOT start a poll loop. The workflow owns polling and will call you back for each event that needs you. Ending your turn does not end the review.",
+			"</instructions>",
+		].join("\n"),
+		"<output_format>Under 150 words: the live review URL, the manual fallback path, how to end the review, and whether the browser opened.</output_format>",
+	].join("\n\n");
+}