@dungle-scrubs/tallow 0.8.27 → 0.9.0

package/skills/tallow-expert/SKILL.md

@@ -34,7 +34,7 @@ Relay that answer to the user.
 | Component | Location |
 |-----------|----------|
 | Core source | `src/` (agent-runner.ts, atomic-write.ts, auth-hardening.ts, cli-auto-rebuild.ts, cli.ts, compaction-cancel-patch.ts, config.ts, extensions-global.d.ts, fatal-errors.ts, index.ts, install.ts, interactive-mode-patch.ts, model-metadata-overrides.ts, otel.ts, pid-manager.ts, pid-schema.ts, plugins.ts, process-cleanup.ts, project-trust-banner.ts, project-trust-interop.ts, project-trust.ts, runtime-path-provider.ts, runtime-provenance.ts, sdk.ts, session-migration.ts, session-utils.ts, startup-profile.ts, startup-timing.ts, streaming-yield-patch.ts, workspace-transition-interactive.ts, workspace-transition-relay.ts, workspace-transition.ts, yield-to-io.ts) |
-| Extensions | `extensions/` — extension.json + index.ts each (52 bundled) |
+| Extensions | `extensions/` — extension.json + index.ts each (53 bundled) |
 | Skills | `skills/` — subdirs with SKILL.md |
 | Agents | `agents/` — markdown with YAML frontmatter |
 | Themes | `themes/` — JSON files (34 dark-only themes) |
@@ -59,8 +59,8 @@ Extensions export a default function receiving `ExtensionAPI` (conventionally na
 
 #### Registration
 
-- `registerTool(tool: ToolDefinition<TParams, TDetails>)` — Register a tool that the LLM can call.
-- `registerCommand(name: string, options: Omit<RegisteredCommand, "name">)` — Register a custom command.
+- `registerTool(tool: ToolDefinition<TParams, TDetails, TState>)` — Register a tool that the LLM can call.
+- `registerCommand(name: string, options: Omit<RegisteredCommand, "name" | "sourceInfo">)` — Register a custom command.
 - `registerFlag(name: string, options: object)` — Register a CLI flag.
 - `registerMessageRenderer(customType: string, renderer: MessageRenderer<T>)` — Register a custom renderer for CustomMessageEntry.
 - `registerProvider(name: string, config: ProviderConfig)` — Register or override a model provider.
@@ -81,7 +81,7 @@ Extensions export a default function receiving `ExtensionAPI` (conventionally na
 - `getFlag(name: string)` — Get the value of a registered CLI flag.
 - `exec(command: string, args: string[], options?: ExecOptions)` — Execute a shell command.
 - `getActiveTools()` — Get the list of currently active tool names.
-- `getAllTools()` — Get all configured tools with name and description.
+- `getAllTools()` — Get all configured tools with parameter schema and source metadata.
 - `setActiveTools(toolNames: string[])` — Set the active tools by name.
 - `getCommands()` — Get available slash commands in the current session.
 - `setModel(model: Model<any>)` — Set the current model.
@@ -154,6 +154,7 @@ Extensions export a default function receiving `ExtensionAPI` (conventionally na
 - `modelRegistry` — Model registry for API key resolution
 - `model` — Current model (may be undefined)
 - `isIdle()` — Whether the agent is idle (not streaming)
+- `signal` — The current abort signal, or undefined when the agent is not streaming.
 - `abort()` — Abort the current agent operation
 - `hasPendingMessages()` — Whether there are queued messages waiting
 - `shutdown()` — Gracefully shutdown pi and exit.
@@ -178,6 +179,7 @@ Extensions export a default function receiving `ExtensionAPI` (conventionally na
 - `onTerminalInput(handler: TerminalInputHandler)` — Listen to raw terminal input (interactive mode only).
 - `setStatus(key: string, text: string)` — Set status text in the footer/status bar.
 - `setWorkingMessage(message?: string)` — Set the working/loading message shown during streaming.
+- `setHiddenThinkingLabel(label?: string)` — Set the label shown for hidden thinking blocks.
 - `setWidget(key: string, content: string[], options?: ExtensionWidgetOptions)` — Set a widget to display above or below the editor.
 - `setTitle(title: string)` — Set the terminal window/tab title.
 - `pasteToEditor(text: string)` — Paste text into the editor, triggering paste handling (collapse for large content).

package/extensions/plan-mode-tool/__tests__/e2e.mjs

@@ -1,350 +0,0 @@
-#!/usr/bin/env node
-
-/**
- * E2E test for the plan-mode extension.
- *
- * Proves:
- *   1. plan_mode tool remains available after toggling modes
- *   2. Plan mode enforces a strict read-only allowlist
- *   3. Non-allowlisted extension tools are blocked in plan mode
- *   4. Disabling plan mode restores normal access
- *
- * Uses the SDK to load ONLY the plan-mode extension (isolated).
- * Costs ~$0.01 per run.
- *
- * Usage:
- *   node extensions/plan-mode-tool/__tests__/e2e.mjs
- */
-
-import fs from "node:fs";
-import os from "node:os";
-import path from "node:path";
-import { fileURLToPath } from "node:url";
-import { getModel } from "@mariozechner/pi-ai";
-import {
-	AuthStorage,
-	createAgentSession,
-	DefaultResourceLoader,
-	ModelRegistry,
-	SessionManager,
-	SettingsManager,
-} from "@mariozechner/pi-coding-agent";
-import { Type } from "@sinclair/typebox";
-
-// ── Helpers ──────────────────────────────────────────────────
-
-const results = [];
-
-/**
- * Record a test result.
- * @param {string} name - Test name
- * @param {boolean} passed - Pass/fail
- * @param {string} [detail] - Extra detail on failure
- */
-function check(name, passed, detail) {
-	results.push({ name, passed, detail });
-	const icon = passed ? "\x1b[32m✓\x1b[0m" : "\x1b[31m✗\x1b[0m";
-	let line = `  ${icon} ${name}`;
-	if (!passed && detail) line += `\n    ${detail.slice(0, 300)}`;
-	console.log(line);
-}
-
-/**
- * Get the text content of the most recent tool result for a given tool name.
- * @param {import("@mariozechner/pi-coding-agent").AgentSession} session
- * @param {string} toolName
- * @returns {string}
- */
-function lastToolResultText(session, toolName) {
-	const msgs = session.messages;
-	for (let i = msgs.length - 1; i >= 0; i--) {
-		const m = msgs[i];
-		if (m.role === "toolResult" && m.toolName === toolName) {
-			for (const part of m.content) {
-				if (part.type === "text") return part.text;
-			}
-		}
-	}
-	return "";
-}
-
-/**
- * Check if any tool result in the session contains "not found" error.
- * @param {import("@mariozechner/pi-coding-agent").AgentSession} session
- * @param {string} toolName
- * @returns {boolean}
- */
-function hasToolNotFoundError(session, toolName) {
-	const msgs = session.messages;
-	for (let i = msgs.length - 1; i >= 0; i--) {
-		const m = msgs[i];
-		if (m.role === "toolResult") {
-			for (const part of m.content) {
-				if (part.type === "text" && part.text.includes(`Tool ${toolName} not found`)) {
-					return true;
-				}
-			}
-		}
-	}
-	return false;
-}
-
-/**
- * Check if a tool call was blocked by plan-mode policy.
- * @param {import("@mariozechner/pi-coding-agent").AgentSession} session
- * @param {string} toolName
- * @returns {boolean}
- */
-function hasPlanModeToolBlockedError(session, toolName) {
-	const msgs = session.messages;
-	for (let i = msgs.length - 1; i >= 0; i--) {
-		const m = msgs[i];
-		if (m.role !== "toolResult") continue;
-		for (const part of m.content) {
-			if (part.type === "text" && part.text.includes(`Plan mode: tool "${toolName}" blocked`)) {
-				return true;
-			}
-		}
-	}
-	return false;
-}
-
-// ── Isolated extension loading ───────────────────────────────
-
-const __dirname = path.dirname(fileURLToPath(import.meta.url));
-const extensionSrcDir = path.resolve(__dirname, "..");
-
-const testAgentDir = fs.mkdtempSync(path.join(os.tmpdir(), "pi-e2e-plan-"));
-const extDst = path.join(testAgentDir, "extensions/plan-mode-tool");
-fs.mkdirSync(extDst, { recursive: true });
-for (const file of ["index.ts", "utils.ts"]) {
-	fs.copyFileSync(path.join(extensionSrcDir, file), path.join(extDst, file));
-}
-
-// ── Setup ────────────────────────────────────────────────────
-
-console.log("\n\x1b[1m══ Plan Mode Extension E2E Test ══\x1b[0m\n");
-
-const authStorage = new AuthStorage();
-const modelRegistry = new ModelRegistry(authStorage);
-const model = getModel("anthropic", "claude-haiku-4-5");
-if (!model) {
-	console.error("✗ Model claude-haiku-4-5 not found");
-	process.exit(1);
-}
-
-const settingsManager = SettingsManager.inMemory({ compaction: { enabled: false } });
-
-/**
- * Register mock tools used to validate strict plan-mode allowlisting.
- * @param {import("@mariozechner/pi-coding-agent").ExtensionAPI} pi
- */
-function registerMockTools(pi) {
-	pi.registerTool({
-		name: "bg_bash",
-		label: "bg_bash",
-		description: "Mock background bash tool",
-		parameters: Type.Object({ command: Type.String() }),
-		async execute(_toolCallId, params) {
-			return {
-				content: [{ type: "text", text: `mock-bg-bash-ok:${params.command}` }],
-				details: {},
-			};
-		},
-	});
-
-	pi.registerTool({
-		name: "subagent",
-		label: "subagent",
-		description: "Mock subagent tool",
-		parameters: Type.Object({ task: Type.String() }),
-		async execute(_toolCallId, params) {
-			return {
-				content: [{ type: "text", text: `mock-subagent-ok:${params.task}` }],
-				details: {},
-			};
-		},
-	});
-
-	pi.registerTool({
-		name: "mcp__mock__ping",
-		label: "mcp__mock__ping",
-		description: "Mock MCP-style tool",
-		parameters: Type.Object({}),
-		async execute() {
-			return {
-				content: [{ type: "text", text: "mock-mcp-ok" }],
-				details: {},
-			};
-		},
-	});
-
-	pi.registerTool({
-		name: "questionnaire",
-		label: "questionnaire",
-		description: "Mock read-only questionnaire tool",
-		parameters: Type.Object({}),
-		async execute() {
-			return {
-				content: [{ type: "text", text: "mock-questionnaire-ok" }],
-				details: {},
-			};
-		},
-	});
-}
-
-console.log("Loading extension (isolated)...");
-const loader = new DefaultResourceLoader({
-	cwd: os.tmpdir(),
-	agentDir: testAgentDir,
-	settingsManager,
-	extensionFactories: [registerMockTools],
-	skillsOverride: () => ({ skills: [], diagnostics: [] }),
-	promptsOverride: () => ({ prompts: [], diagnostics: [] }),
-	agentsFilesOverride: () => ({ agentsFiles: [] }),
-});
-await loader.reload();
-
-const exts = loader.getExtensions();
-console.log(`  Extensions loaded: ${exts.extensions.length}, errors: ${exts.errors.length}`);
-if (exts.errors.length > 0) {
-	console.error("  Extension errors:", exts.errors);
-}
-
-console.log("Creating session (haiku)...\n");
-const { session } = await createAgentSession({
-	model,
-	thinkingLevel: "off",
-	authStorage,
-	modelRegistry,
-	resourceLoader: loader,
-	sessionManager: SessionManager.inMemory(),
-	settingsManager,
-});
-
-// Log tool calls
-session.subscribe((event) => {
-	if (event.type === "tool_execution_start") {
-		process.stdout.write(`    \x1b[2m→ ${event.toolName}\x1b[0m\n`);
-	}
-});
-
-// ── Test 1: plan_mode tool exists at startup ─────────────────
-
-console.log("\x1b[1mTest 1: plan_mode tool available at startup\x1b[0m");
-await session.prompt(
-	'Call the plan_mode tool with action "status". Only call this one tool, nothing else.'
-);
-const statusText = lastToolResultText(session, "plan_mode");
-const noStartupError = !hasToolNotFoundError(session, "plan_mode");
-check("plan_mode tool callable at startup", noStartupError, statusText);
-check("reports normal mode", statusText.includes("normal"), statusText);
-
-// ── Test 2: Enable plan mode, verify plan_mode survives ──────
-
-console.log("\n\x1b[1mTest 2: Enable plan mode → plan_mode tool still available\x1b[0m");
-await session.prompt(
-	'Call the plan_mode tool with action "enable". Only call this one tool, nothing else.'
-);
-const enableText = lastToolResultText(session, "plan_mode");
-const noEnableError = !hasToolNotFoundError(session, "plan_mode");
-check("plan_mode callable during enable", noEnableError, enableText);
-check("reports plan mode enabled", enableText.includes("enabled"), enableText);
-
-// Now check status — plan_mode should still work IN plan mode
-await session.prompt(
-	'Call the plan_mode tool with action "status". Only call this one tool, nothing else.'
-);
-const planStatusText = lastToolResultText(session, "plan_mode");
-const noPlanStatusError = !hasToolNotFoundError(session, "plan_mode");
-check("plan_mode callable while in plan mode", noPlanStatusError, planStatusText);
-check("reports planning mode", planStatusText.includes("planning"), planStatusText);
-
-// ── Test 3: Disable plan mode, verify plan_mode survives ─────
-
-console.log("\n\x1b[1mTest 3: Disable plan mode → plan_mode tool still available\x1b[0m");
-await session.prompt(
-	'Call the plan_mode tool with action "disable". Only call this one tool, nothing else.'
-);
-const disableText = lastToolResultText(session, "plan_mode");
-const noDisableError = !hasToolNotFoundError(session, "plan_mode");
-check("plan_mode callable during disable", noDisableError, disableText);
-check("reports disabled", disableText.includes("disabled"), disableText);
-
-// Final status check — should be back to normal
-await session.prompt(
-	'Call the plan_mode tool with action "status". Only call this one tool, nothing else.'
-);
-const finalStatusText = lastToolResultText(session, "plan_mode");
-const noFinalError = !hasToolNotFoundError(session, "plan_mode");
-check("plan_mode callable after round-trip", noFinalError, finalStatusText);
-check("back to normal mode", finalStatusText.includes("normal"), finalStatusText);
-
-// ── Test 4: Strict allowlist enforcement in plan mode ────────
-
-console.log("\n\x1b[1mTest 4: Strict allowlist blocks non-read-only tools\x1b[0m");
-await session.prompt(
-	'Call the plan_mode tool with action "enable". Only call this one tool, nothing else.'
-);
-
-await session.prompt(
-	'Call the edit tool to edit file "/tmp/test.txt" replacing "a" with "b". Only call edit, nothing else.'
-);
-const editBlocked =
-	hasToolNotFoundError(session, "edit") || hasPlanModeToolBlockedError(session, "edit");
-check("edit tool blocked in plan mode", editBlocked, "edit should not be available in plan mode");
-
-await session.prompt(
-	'Call the bg_bash tool with command "echo blocked". Only call bg_bash, nothing else.'
-);
-const bgBashBlocked =
-	hasToolNotFoundError(session, "bg_bash") || hasPlanModeToolBlockedError(session, "bg_bash");
-check("bg_bash blocked in plan mode", bgBashBlocked, "bg_bash should be blocked in plan mode");
-
-await session.prompt('Call the subagent tool with task "ping". Only call subagent, nothing else.');
-const subagentBlocked =
-	hasToolNotFoundError(session, "subagent") || hasPlanModeToolBlockedError(session, "subagent");
-check("subagent blocked in plan mode", subagentBlocked, "subagent should be blocked in plan mode");
-
-await session.prompt("Call the mcp__mock__ping tool. Only call this one tool, nothing else.");
-const mcpBlocked =
-	hasToolNotFoundError(session, "mcp__mock__ping") ||
-	hasPlanModeToolBlockedError(session, "mcp__mock__ping");
-check("mcp__* tools blocked in plan mode", mcpBlocked, "MCP tools should be blocked in plan mode");
-
-await session.prompt("Call the questionnaire tool. Only call this one tool, nothing else.");
-const questionnaireText = lastToolResultText(session, "questionnaire");
-const questionnaireAllowed = questionnaireText.includes("mock-questionnaire-ok");
-check("allowlisted questionnaire tool still works", questionnaireAllowed, questionnaireText);
-
-// ── Test 5: Disabling plan mode restores normal access ───────
-
-console.log("\n\x1b[1mTest 5: Disable restores normal tool access\x1b[0m");
-await session.prompt(
-	'Call the plan_mode tool with action "disable". Only call this one tool, nothing else.'
-);
-await session.prompt(
-	'Call the subagent tool with task "after-disable". Only call subagent, nothing else.'
-);
-const subagentAfterDisableText = lastToolResultText(session, "subagent");
-const subagentRestored = subagentAfterDisableText.includes("mock-subagent-ok:after-disable");
-check("subagent restored after disabling plan mode", subagentRestored, subagentAfterDisableText);
-
-// ── Cleanup & Summary ────────────────────────────────────────
-
-session.dispose();
-fs.rmSync(testAgentDir, { recursive: true, force: true });
-
-const passed = results.filter((r) => r.passed).length;
-const total = results.length;
-
-console.log(`\n\x1b[1m══ Results: ${passed}/${total} passed ══\x1b[0m`);
-if (passed < total) {
-	console.log("\n\x1b[31mFailed:\x1b[0m");
-	for (const r of results.filter((r) => !r.passed)) {
-		console.log(`  ✗ ${r.name}`);
-		if (r.detail) console.log(`    ${r.detail.slice(0, 300)}`);
-	}
-}
-console.log();
-process.exit(passed === total ? 0 : 1);

package/extensions/plan-mode-tool/__tests__/index.test.ts

@@ -1,213 +0,0 @@
-import { beforeEach, describe, expect, test } from "bun:test";
-import type { ExtensionAPI, ExtensionContext, ToolDefinition } from "@mariozechner/pi-coding-agent";
-import { Type } from "@sinclair/typebox";
-import { ExtensionHarness } from "../../../test-utils/extension-harness.js";
-import planModeExtension from "../index.js";
-import { PLAN_MODE_ALLOWED_TOOLS } from "../utils.js";
-
-const BASELINE_TOOLS = [
-	"read",
-	"bash",
-	"grep",
-	"find",
-	"ls",
-	"edit",
-	"write",
-	"subagent",
-	"bg_bash",
-	"mcp__mock__ping",
-	"questionnaire",
-	"plan_mode",
-] as const;
-
-/**
- * Register mock tools used to test plan-mode gating and restoration.
- *
- * @param pi - Extension API test double
- * @returns void
- */
-function registerMockTools(pi: ExtensionAPI): void {
-	const names = [
-		"read",
-		"bash",
-		"grep",
-		"find",
-		"ls",
-		"edit",
-		"write",
-		"subagent",
-		"bg_bash",
-		"mcp__mock__ping",
-		"questionnaire",
-	] as const;
-
-	for (const name of names) {
-		pi.registerTool({
-			name,
-			label: name,
-			description: `Mock ${name}`,
-			parameters: Type.Object({}),
-			async execute() {
-				return {
-					content: [{ type: "text", text: `${name}-ok` }],
-					details: {},
-				};
-			},
-		});
-	}
-}
-
-/**
- * Create an extension context with optional persisted session entries.
- *
- * @param entries - Session entries returned by sessionManager.getEntries
- * @returns Context object compatible with extension handlers
- */
-function createContext(entries: unknown[] = [], hasUI = true): ExtensionContext {
-	return {
-		cwd: process.cwd(),
-		hasUI,
-		ui: {
-			notify() {},
-			setStatus() {},
-			setEditorComponent() {},
-			setWidget() {},
-			theme: {
-				fg(_token: string, value: string) {
-					return value;
-				},
-				strikethrough(value: string) {
-					return value;
-				},
-			},
-		} as never,
-		sessionManager: {
-			getEntries() {
-				return entries;
-			},
-		} as never,
-	} as unknown as ExtensionContext;
-}
-
-/**
- * Resolve a registered tool from the test harness.
- *
- * @param harness - Extension harness
- * @param name - Tool name
- * @returns Tool definition
- */
-function getTool(harness: ExtensionHarness, name: string): ToolDefinition {
-	const tool = harness.tools.get(name);
-	if (!tool) throw new Error(`Tool not registered: ${name}`);
-	return tool;
-}
-
-describe("plan-mode strict readonly enforcement", () => {
-	let harness: ExtensionHarness;
-
-	beforeEach(async () => {
-		harness = ExtensionHarness.create();
-		await harness.loadExtension(registerMockTools);
-		await harness.loadExtension(planModeExtension);
-		harness.api.setActiveTools([...BASELINE_TOOLS]);
-	});
-
-	test("enable applies strict allowlist and disable restores previous tools", async () => {
-		const tool = getTool(harness, "plan_mode");
-		const ctx = createContext();
-
-		await tool.execute("tc-enable", { action: "enable" }, undefined, () => {}, ctx);
-		expect(harness.api.getActiveTools()).toEqual(
-			PLAN_MODE_ALLOWED_TOOLS.filter((name) => BASELINE_TOOLS.includes(name))
-		);
-
-		await tool.execute("tc-disable", { action: "disable" }, undefined, () => {}, ctx);
-		expect(harness.api.getActiveTools()).toEqual([...BASELINE_TOOLS]);
-	});
-
-	test("tool_call blocks non-allowlisted tools and unsafe bash", async () => {
-		const tool = getTool(harness, "plan_mode");
-		const ctx = createContext();
-		await tool.execute("tc-enable", { action: "enable" }, undefined, () => {}, ctx);
-
-		const [blockedToolResult] = await harness.fireEvent(
-			"tool_call",
-			{ toolName: "subagent", input: { task: "x" } },
-			ctx
-		);
-		expect(blockedToolResult).toMatchObject({ block: true });
-		expect((blockedToolResult as { reason: string }).reason).toContain('tool "subagent" blocked');
-
-		const [safeBashResult] = await harness.fireEvent(
-			"tool_call",
-			{ toolName: "bash", input: { command: "ls -la" } },
-			ctx
-		);
-		expect(safeBashResult).toBeUndefined();
-
-		const [unsafeBashResult] = await harness.fireEvent(
-			"tool_call",
-			{ toolName: "bash", input: { command: "rm -rf /tmp/nope" } },
-			ctx
-		);
-		expect(unsafeBashResult).toMatchObject({ block: true });
-	});
-
-	test("resumed plan mode re-applies strict policy", async () => {
-		const persistedEntries = [
-			{
-				type: "custom",
-				customType: "plan-mode",
-				data: {
-					enabled: true,
-					normalTools: [...BASELINE_TOOLS],
-					todos: [],
-				},
-			},
-		];
-		const ctx = createContext(persistedEntries);
-
-		await harness.fireEvent("session_start", { type: "session_start" }, ctx);
-
-		expect(harness.api.getActiveTools()).toEqual(
-			PLAN_MODE_ALLOWED_TOOLS.filter((name) => BASELINE_TOOLS.includes(name))
-		);
-
-		const [blockedResult] = await harness.fireEvent(
-			"tool_call",
-			{ toolName: "bg_bash", input: { command: "echo hi" } },
-			ctx
-		);
-		expect(blockedResult).toMatchObject({ block: true });
-	});
-
-	test("auto-enable only triggers for interactive UI input", async () => {
-		const [result] = await harness.fireEvent(
-			"input",
-			{ source: "interactive", text: "plan only fix auth" },
-			createContext([], true)
-		);
-
-		expect(result).toEqual({ action: "transform", text: "fix auth" });
-		expect(harness.api.getActiveTools()).toEqual(
-			PLAN_MODE_ALLOWED_TOOLS.filter((name) => BASELINE_TOOLS.includes(name))
-		);
-	});
-
-	test("auto-enable ignores headless or non-interactive input", async () => {
-		const [headlessResult] = await harness.fireEvent(
-			"input",
-			{ source: "interactive", text: "plan only fix auth" },
-			createContext([], false)
-		);
-		const [rpcResult] = await harness.fireEvent(
-			"input",
-			{ source: "rpc", text: "plan only fix auth" },
-			createContext([], true)
-		);
-
-		expect(headlessResult).toEqual({ action: "continue" });
-		expect(rpcResult).toEqual({ action: "continue" });
-		expect(harness.api.getActiveTools()).toEqual([...BASELINE_TOOLS]);
-	});
-});