pi-ui-extend 1.0.19 → 1.0.20

package/external/pi-tools-suite/README.md

@@ -251,8 +251,9 @@ The built-in `browser-qa` role runs on `antigravity/gemini-3-flash-preview`, wit
 `openai-codex/gpt-5.4-mini` as its fallback. Its browser workflow is an explicit
 private skill under `src/async-subagents/private-skills/`, outside normal Pi skill
 discovery. The role's first-class `isolatedSkills` setting launches the child with
-`--no-skills` plus only that skill, so parent and ordinary sub-agent sessions do
-not discover it.
+`--no-skills` plus the private workflow and the packaged `skills/playwright-cli`
+skill. Both are mandatory even when configuration adds other isolated skills;
+parent and ordinary sub-agent sessions do not discover the private workflow.
 
 Keep named dev/staging auth profiles in project `.pi/qa_auth.jsonc` (there is no
 `/qa-auth` command). The private runner supports `form`, `cookie`, `localStorage`,

package/external/pi-tools-suite/docs/browser-qa-subagent.md

@@ -17,6 +17,9 @@ evidence. The role uses `antigravity/gemini-3-flash-preview`, falling back to
   Antigravity-backed role unavailable.
 - A type profile may declare `isolatedSkills`. Spawning that profile adds
   `--no-skills` followed by one explicit `--skill` per configured path.
+- The `browser-qa` profile always loads both its private workflow and the
+  packaged `skills/playwright-cli` skill. Configuration may append isolated
+  skills but cannot remove either mandatory browser skill.
 - Other sub-agent profiles and the parent session must not discover the private
   skill automatically.
 
@@ -66,9 +69,10 @@ evidence. The role uses `antigravity/gemini-3-flash-preview`, falling back to
 
 ## Acceptance criteria
 
-1. `browser-qa` resolves to the intended model/fallback and private skill, and
-   its isolated child process can register the configured Antigravity model.
-2. Spawn args contain `--no-skills` and only the explicit private skill for this
+1. `browser-qa` resolves to the intended model/fallback, private workflow, and
+   packaged `playwright-cli` skill, and its isolated child process can register
+   the configured Antigravity model.
+2. Spawn args contain `--no-skills` and both mandatory explicit skills for this
    profile; ordinary profiles retain existing skill discovery behavior.
 3. Auth profile listing and all error output are redacted; model-authored input
    cannot execute code in the credential-bearing process.
@@ -79,3 +83,19 @@ evidence. The role uses `antigravity/gemini-3-flash-preview`, falling back to
 5. Completed test runs report clickable screenshot, video, and trace links
    whenever those artifacts exist.
 6. Suite tests/typecheck, host checks, and suite sync pass.
+
+## Real-browser regression test
+
+The repository includes a local mock-page E2E that launches real Chromium and
+asserts PNG screenshots, WebM video, sanitized trace output, and absolute
+path/`file:` URI metadata:
+
+```bash
+npx playwright install chromium
+npm run test:browser-qa-e2e
+```
+
+Normal suite tests keep this case skipped; the Publish workflow runs it on
+Linux after installing Chromium. Explicit runs retain the latest artifacts in
+`.pi/qa-runs/browser-qa-e2e/latest/` and print a clickable link for every file.
+Set `BROWSER_QA_KEEP_EVIDENCE=0` to clean up without publishing evidence.

package/external/pi-tools-suite/package.json

@@ -22,6 +22,7 @@
     "smoke:tools": "PI_OFFLINE=1 pi --no-session -p \"ping\"",
     "smoke": "npm run smoke:explicit && npm run smoke:auto && npm run smoke:tools",
     "test": "bun test test",
+    "test:browser-qa-e2e": "BROWSER_QA_RUNNER_E2E=1 bun test test/async-subagents/browser-qa-runner.e2e.test.ts",
     "test:async-subagents-e2e": "ASYNC_SUBAGENTS_E2E=1 ASYNC_SUBAGENTS_DEBUG_LOGS=1 ASYNC_SUBAGENTS_MODEL=zai/glm-5-turbo bun test --concurrent --max-concurrency=30 test/async-subagents",
     "test:async-subagents-selection-e2e": "ASYNC_SUBAGENTS_SELECTION_E2E=1 ASYNC_SUBAGENTS_MODEL=zai/glm-5-turbo bun test --concurrent --max-concurrency=30 test/async-subagents/selection-e2e.test.ts",
     "test:prompt-evals:tool-selection": "PROMPT_EVAL_E2E=1 bun test --concurrent --max-concurrency=10 test/tool-selection-e2e.test.ts",

package/external/pi-tools-suite/src/async-subagents/core/config.ts

@@ -288,6 +288,21 @@ export function getBrowserQaSkillPath(): string {
 	return path.resolve(path.dirname(fileURLToPath(import.meta.url)), "..", "private-skills", "browser-qa", "SKILL.md");
 }
 
+export function getPlaywrightCliSkillPath(): string {
+	const bundledPath = path.resolve(
+		path.dirname(fileURLToPath(import.meta.url)),
+		"../../../../..",
+		"skills",
+		"playwright-cli",
+		"SKILL.md",
+	);
+	const installedPaths = [
+		path.join(os.homedir(), ".agents", "skills", "playwright-cli", "SKILL.md"),
+		path.join(os.homedir(), ".pi", "agent", "skills", "playwright-cli", "SKILL.md"),
+	];
+	return [bundledPath, ...installedPaths].find((candidate) => fs.existsSync(candidate)) ?? bundledPath;
+}
+
 export function getSubagentConfigInitTargetPath(cwd: string, env: NodeJS.ProcessEnv = process.env): string {
 	return explicitSubagentConfigPath(cwd, env) ?? getDefaultSubagentConfigPath();
 }
@@ -381,7 +396,7 @@ export function resolveAgentTaskConfig(
 	return {
 		profile,
 		extraArgs,
-		isolatedSkills: arrayOfStrings(profile?.isolatedSkills) ?? [],
+		isolatedSkills: resolveIsolatedSkills(selectedType, profile),
 		fallbackModels,
 		retry: resolveRetryConfig(config.retry, profile?.retry),
 		maxResultBytes: profile?.maxResultBytes ?? config.maxResultBytes,
@@ -399,6 +414,12 @@ export function resolveAgentTaskConfig(
 	};
 }
 
+function resolveIsolatedSkills(selectedType: string | undefined, profile: SubagentTypeConfig | undefined): string[] {
+	const configured = arrayOfStrings(profile?.isolatedSkills) ?? [];
+	if (selectedType !== "browser-qa") return configured;
+	return [...new Set([getBrowserQaSkillPath(), getPlaywrightCliSkillPath(), ...configured])];
+}
+
 export function resolveSubagentRoutingConfig(config: SubagentConfig): ResolvedSubagentRoutingConfig {
 	return { ...DEFAULT_ROUTING_CONFIG, ...(config.routing ?? {}) };
 }

package/external/pi-tools-suite/src/async-subagents/lib.ts

@@ -24,6 +24,7 @@ export {
 	existingSubagentConfigFiles,
 	getDefaultSubagentConfigPath,
 	getBrowserQaSkillPath,
+	getPlaywrightCliSkillPath,
 	getSubagentConfigInitTargetPath,
 	getSubagentConfigSamplePath,
 	isBlindModelRef,

package/external/pi-tools-suite/src/todo/state/replay.ts

@@ -1,6 +1,8 @@
 import type { TaskDetails } from "../tool/types.js";
 import { EMPTY_STATE, type TaskState } from "./state.js";
 
+export const TODO_STATE_ENTRY_TYPE = "pi-tools-suite:todo-state";
+
 /**
  * Discriminator for `details` envelopes that match the persisted `TaskDetails`
  * shape. Defensive — branch entries from older or corrupt sessions are
@@ -12,10 +14,29 @@ export function isTaskDetails(value: unknown): value is TaskDetails {
 	return Array.isArray(v.tasks) && typeof v.nextId === "number";
 }
 
+function taskDetailsFromEntry(entry: unknown): TaskDetails | undefined {
+	const candidate = entry as {
+		type?: string;
+		customType?: string;
+		data?: unknown;
+		message?: { role?: string; toolName?: string; details?: unknown };
+	};
+
+	if (candidate.type === "custom" && candidate.customType === TODO_STATE_ENTRY_TYPE) {
+		return isTaskDetails(candidate.data) ? candidate.data : undefined;
+	}
+
+	const message = candidate.message;
+	if (candidate.type !== "message" || message?.role !== "toolResult" || message.toolName !== "todo") return undefined;
+	return isTaskDetails(message.details) ? message.details : undefined;
+}
+
 /**
- * Walk the current branch in chronological order; the LAST `toolResult` whose
- * `toolName === "todo"` and whose `details` shape matches `TaskDetails` wins
- * (last-write-wins). When no matching entry exists, returns `EMPTY_STATE`.
+ * Walk the current branch in chronological order; the LAST valid snapshot
+ * from either a `todo` tool result or a todo custom state entry wins
+ * (last-write-wins). Custom entries persist slash-command mutations, which do
+ * not otherwise produce tool results. When no matching entry exists, returns
+ * `EMPTY_STATE`.
  *
  * Pure of module state — `index.ts` writes the returned snapshot into the
  * store after this returns. The function explicitly does NOT touch the store
@@ -24,14 +45,11 @@ export function isTaskDetails(value: unknown): value is TaskDetails {
 export function replayFromBranch(ctx: { sessionManager: { getBranch(): Iterable<unknown> } }): TaskState {
 	let result: TaskState = { tasks: [...EMPTY_STATE.tasks], nextId: EMPTY_STATE.nextId };
 	for (const entry of ctx.sessionManager.getBranch()) {
-		const e = entry as { type?: string; message?: { role?: string; toolName?: string; details?: unknown } };
-		if (e.type !== "message") continue;
-		const msg = e.message;
-		if (!msg || msg.role !== "toolResult" || msg.toolName !== "todo") continue;
-		if (!isTaskDetails(msg.details)) continue;
+		const details = taskDetailsFromEntry(entry);
+		if (!details) continue;
 		result = {
-			tasks: msg.details.tasks.map((t) => ({ ...t })),
-			nextId: msg.details.nextId,
+			tasks: details.tasks.map((t) => ({ ...t })),
+			nextId: details.nextId,
 		};
 	}
 	return result;

package/external/pi-tools-suite/src/todo/todo.ts

@@ -23,7 +23,7 @@ import {
 	syncPersistedPlan,
 } from "./state/persistence.js";
 import { AUTO_CLEAR_COMPLETED_MESSAGE, autoClearCompletedTodos } from "./state/auto-clear.js";
-import { replayFromBranch } from "./state/replay.js";
+import { replayFromBranch, TODO_STATE_ENTRY_TYPE } from "./state/replay.js";
 import { isTaskBlocked, selectTasksByStatus, selectTodoCounts } from "./state/selectors.js";
 import { applyTaskMutation } from "./state/state-reducer.js";
 import { activateStateScope, commitState, getState, replaceState } from "./state/store.js";
@@ -98,6 +98,7 @@ interface TodoToolRegistrationOptions extends TodoToolHooks {
 
 type TodoStateEventContext = { sessionManager?: { getSessionFile?: () => unknown; getSessionId?: () => unknown } };
 type TodoStateEventEmitter = { events?: { emit?: (channel: string, data: unknown) => void } };
+type TodoStateEntryWriter = { appendEntry<T = unknown>(customType: string, data?: T): void };
 
 interface TodosCommandOptions {
 	status?: TaskStatus;
@@ -238,6 +239,20 @@ export function publishTodoState(
 	});
 }
 
+function appendTodoStateSnapshot(
+	pi: TodoStateEntryWriter,
+	action: TaskAction,
+	params: Record<string, unknown>,
+): void {
+	const state = getState();
+	pi.appendEntry(TODO_STATE_ENTRY_TYPE, {
+		action,
+		params,
+		tasks: state.tasks,
+		nextId: state.nextId,
+	});
+}
+
 function handlePersistCommand(args: unknown, ctx: { cwd?: string; hasUI?: boolean; ui?: { notify?: (message: string, level?: NotifyLevel) => void } }): boolean {
 	const [command, subcommand = "status"] = getCommandTokens(args);
 	if (command !== "persist") return false;
@@ -310,12 +325,13 @@ function handleScopeCommand(
 }
 
 function clearTodos(
-	pi: TodoStateEventEmitter,
+	pi: TodoStateEventEmitter & TodoStateEntryWriter,
 	ctx: { cwd?: string; hasUI?: boolean; ui?: { notify?: (message: string, level?: NotifyLevel) => void } },
 ): void {
 	const result = applyTaskMutation(getState(), "clear", { action: "clear" });
 	if (result.op.kind !== "clear") return;
 	commitState(result.state);
+	appendTodoStateSnapshot(pi, "clear", { action: "clear" });
 	publishTodoState(pi, ctx, "clear", { action: "clear" });
 	const sync = syncPersistedPlan(ctx.cwd, result.state);
 	const persistedText = sync?.completed ? `\nProject todo plan removed: ${sync.path}` : "";
@@ -449,9 +465,12 @@ export function registerTodosCommand(pi: ExtensionAPI): void {
 		handler: async (args, ctx) => {
 			activateTodoStateScope(ctx);
 			if (handlePersistCommand(args, ctx)) return;
-			if (handleScopeCommand(args, ctx, () => publishTodoState(pi as TodoStateEventEmitter, ctx))) return;
+			if (handleScopeCommand(args, ctx, () => {
+				appendTodoStateSnapshot(pi, "update", { command: "scope" });
+				publishTodoState(pi as TodoStateEventEmitter, ctx);
+			})) return;
 			if (getCommandTokens(args)[0] === "clear") {
-				clearTodos(pi as TodoStateEventEmitter, ctx);
+				clearTodos(pi, ctx);
 				return;
 			}
 			if (!ctx.hasUI) {
@@ -529,7 +548,10 @@ export function registerTodosCommand(pi: ExtensionAPI): void {
 		description: "Select todo ids to continue from a persisted plan; pending/in_progress items outside the scope become deferred.",
 		handler: async (args, ctx) => {
 			activateTodoStateScope(ctx);
-			handleScopeCommand(`scope ${getCommandText(args)}`, ctx, () => publishTodoState(pi as TodoStateEventEmitter, ctx));
+			handleScopeCommand(`scope ${getCommandText(args)}`, ctx, () => {
+				appendTodoStateSnapshot(pi, "update", { command: "scope" });
+				publishTodoState(pi as TodoStateEventEmitter, ctx);
+			});
 		},
 	});
 
@@ -537,7 +559,7 @@ export function registerTodosCommand(pi: ExtensionAPI): void {
 		description: "Clear all todos on the current branch and remove the persisted project plan if enabled.",
 		handler: async (_args, ctx) => {
 			activateTodoStateScope(ctx);
-			clearTodos(pi as TodoStateEventEmitter, ctx);
+			clearTodos(pi, ctx);
 		},
 	});
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-ui-extend",
-  "version": "1.0.19",
+  "version": "1.0.20",
   "description": "Pix: a workspace-first terminal UI for Pi with tabs, readable tool activity, voice input, and bundled agent tools.",
   "private": false,
   "repository": {
@@ -58,6 +58,7 @@
     "test": "mise exec node@24.16.0 -- npm run test:inner",
     "test:inner": "node --import tsx --test \"tests/**/*.test.ts\"",
     "test:tools-suite": "npm --prefix external/pi-tools-suite test",
+    "test:browser-qa-e2e": "npm --prefix external/pi-tools-suite run test:browser-qa-e2e",
     "sync:pi-tools-suite": "node scripts/sync-pi-tools-suite.mjs",
     "sync:pi-tools-suite:check": "node scripts/sync-pi-tools-suite.mjs --check",
     "sync:sdk-pin": "node scripts/sync-sdk-pin.mjs",
@@ -86,6 +87,7 @@
   },
   "devDependencies": {
     "@types/node": "24.12.4",
+    "playwright": "1.62.1",
     "tsx": "4.22.1",
     "typescript": "5.9.3"
   },