feat: add answer extraction replay commands

README.md

@@ -36,7 +36,11 @@ Once installed, this package gives the agent a native way to ask for clarificati
 - 👀 Review tab with `Submit`, `Elaborate`, and `Cancel`
 - 💬 Elaboration flow to capture note-based clarification before final submission
 - ⌨️ Fast keyboard-first interaction with customizable ask keymaps (also mobile-friendly in remote sessions)
-- ⚙️ Ask settings with persisted behaviour and keymap customization via config
+- ⚙️ Ask settings with persisted behaviour, keymap customization, and `/answer` extraction config
+- 🔁 Slash commands for fallback/replay:
+  - `/answer` extracts questions from the latest assistant message into an ask flow
+  - `/answer:again` reopens the latest `/answer` form on the current branch
+  - `/ask:replay` replays the latest real `ask_user` form on the current branch
 - 🗣️ You can talk to your agent to configure pi-ask; it will read the bundled configuration guide and tailor the config for you
 
 ## Feature walkthrough
@@ -129,7 +133,16 @@ You can edit this file yourself, ask pi to edit it for you, or use `/ask-setting
 
 ```json
 {
-  "schemaVersion": 1,
+  "schemaVersion": 2,
+  "answer": {
+    "extractionModels": [
+      { "provider": "openai-codex", "id": "gpt-5.4-mini" },
+      { "provider": "github-copilot", "id": "gpt-5.4-mini" },
+      { "provider": "anthropic", "id": "claude-haiku-4-5" }
+    ],
+    "extractionTimeoutMs": 30000,
+    "extractionRetries": 1
+  },
   "behaviour": {
     "autoSubmitWhenAnsweredWithoutNotes": false,
     "confirmDismissWhenDirty": true,
@@ -151,10 +164,21 @@ Accepted notation follows pi-tui key ids. Common aliases are normalized, for exa
 
 ## Use
 
-After installation, the extension registers the `ask_user` tool and `/ask-settings` command.
+After installation, the extension registers the `ask_user` tool plus `/ask-settings`, `/answer`, `/answer:again`, and `/ask:replay` commands.
 
 Agents can auto-discover and call `ask_user` when they need clarification instead of guessing. In interactive sessions, it opens a terminal UI flow for structured answers, supports native pi-style `@` file references while typing answers or notes, and returns normalized answers back to the agent. Ask settings are available both from `?` in the ask flow and from the `/ask-settings` command. Behaviour settings are binary `on`/`off` toggles that save immediately; customizable ask keymaps are changed by editing the shown config file path.
 
+### Answer and replay commands
+
+`/answer` is useful when the agent asked questions in plain text instead of using `ask_user`. It extracts questions from the latest completed assistant message and opens the same ask UI.
+
+Replay commands are branch-aware. They read persisted entries from the current pi session branch, so they work naturally with `/resume`, `/tree`, and conversation branching:
+
+- `/answer:again` reopens the latest form created by `/answer` on this branch
+- `/ask:replay` reopens the latest real `ask_user` form on this branch
+
+Cancellation is local to the UI: closing a replayed form does not start a new agent turn. Submitted answers are sent back as a normal user follow-up message.
+
 You can also talk to pi to configure this extension. When asked to customize pi-ask settings or keymaps, the agent is instructed to read the bundled `docs/configuration.md` guide first and then edit the config file accordingly.
 
 This package also bundles the `ask-user` skill profile from `skills/ask-user/SKILL.md`. It reinforces when to use the tool, is enabled by default when installed, and can be disabled via `pi config`. The skill was inspired by https://github.com/edlsh/pi-ask-user.

docs/README.md

@@ -4,7 +4,7 @@ This folder keeps only the documentation needed to understand and maintain the e
 
 ## Files
 
-- `configuration.md` — source of truth for configuring pi-ask keymaps and behaviour
+- `configuration.md` — source of truth for configuring pi-ask keymaps, behaviour, and `/answer` extraction
 - `contract.md` — external behavior, full tool payload/result details, and UX guarantees
 - `architecture.md` — module boundaries and invariants
 - `../skills/ask-user/SKILL.md` — auto-bundled agent-side decision-gate guidance; enabled by default, but can be disabled via `pi config`; inspired by https://github.com/edlsh/pi-ask-user

docs/architecture.md

@@ -14,7 +14,10 @@ The codebase is split so the implementation reads through file boundaries and na
 ### Tool surface
 
 - `src/index.ts` — extension entrypoint
-- `src/ask-tool.ts` — tool registration, non-interactive fallback, transcript rendering
+- `src/ask-tool.ts` — tool registration, non-interactive fallback, transcript rendering, ask payload capture
+- `src/answer-commands.ts` — `/answer`, `/answer:again`, and `/ask:replay` command wiring
+- `src/answer-extraction.ts` — configured extraction model selection and raw-JSON extraction retries
+- `src/ask-payload-store.ts` — branch-aware persisted ask payload lookup
 - `src/schema.ts` — TypeBox schema
 - `src/types.ts` — shared types
 
@@ -72,6 +75,7 @@ The codebase is split so the implementation reads through file boundaries and na
 - editor lifecycle stays in the controller, not in the reducers
 - persisted ask settings are migrated to the current schema version, validated, and normalized before use
 - config schema migrations preserve user-provided values and add new defaults only when fields are absent
+- replay payload lookup scans only the current session branch and revalidates payloads before use
 - invalid persisted keymaps fall back to default keymaps for the current session without discarding valid behaviour settings
 - ask settings behaviour changes persist immediately from the settings list
 - when the ask config file is missing, the first ask use writes a default persisted config snapshot under `~/.pi/agent/extensions/`

docs/configuration.md

@@ -28,7 +28,16 @@ Unsupported future versions or invalid files are backed up and defaults are load
 
 ```json
 {
-  "schemaVersion": 1,
+  "schemaVersion": 2,
+  "answer": {
+    "extractionModels": [
+      { "provider": "openai-codex", "id": "gpt-5.4-mini" },
+      { "provider": "github-copilot", "id": "gpt-5.4-mini" },
+      { "provider": "anthropic", "id": "claude-haiku-4-5" }
+    ],
+    "extractionTimeoutMs": 30000,
+    "extractionRetries": 1
+  },
   "behaviour": {
     "autoSubmitWhenAnsweredWithoutNotes": false,
     "confirmDismissWhenDirty": true,
@@ -46,6 +55,29 @@ Unsupported future versions or invalid files are backed up and defaults are load
 }
 ```
 
+## Answer extraction
+
+These settings affect only the `/answer` command. Normal `ask_user` tool calls do not use an extraction model.
+
+### `answer.extractionModels`
+
+- type: array of `{ "provider": string, "id": string }`
+- default: `openai-codex/gpt-5.4-mini`, `github-copilot/gpt-5.4-mini`, then `anthropic/claude-haiku-4-5`
+- effect: `/answer` tries configured models in order and uses the first model with available auth
+- fallback: if no configured model is usable, `/answer` tries the current chat model after validating its auth
+
+### `answer.extractionTimeoutMs`
+
+- type: positive number
+- default: `30000`
+- effect: per-attempt extraction timeout in milliseconds
+
+### `answer.extractionRetries`
+
+- type: integer from `0` to `3`
+- default: `1`
+- effect: number of retry attempts after raw JSON parsing fails; retries include the parse error and previous response as feedback
+
 ## Behaviour
 
 ### `behaviour.autoSubmitWhenAnsweredWithoutNotes`
@@ -184,7 +216,14 @@ Invalid keymaps include:
 
 ```json
 {
-  "schemaVersion": 1,
+  "schemaVersion": 2,
+  "answer": {
+    "extractionRetries": 1,
+    "extractionTimeoutMs": 30000,
+    "extractionModels": [
+      { "provider": "openai-codex", "id": "gpt-5.4-mini" }
+    ]
+  },
   "behaviour": {
     "autoSubmitWhenAnsweredWithoutNotes": true,
     "confirmDismissWhenDirty": true,
@@ -207,7 +246,9 @@ Invalid keymaps include:
 When editing this config for a user:
 
 - preserve unrelated fields
-- keep `schemaVersion` at `1`
+- keep `schemaVersion` at `2`
+- preserve `answer.extractionModels` as explicit provider/id pairs
+- keep `answer.extractionRetries` between `0` and `3`
 - do not assign reserved bindings to configurable actions
 - do not create duplicate configurable bindings
 - after changing the file, tell the user to run `/reload` or restart pi

docs/contract.md

@@ -200,6 +200,10 @@ This document defines the stable external behavior. It does not explain internal
 - on the review tab, `Submit` and `Cancel` preview notes only for answered questions
 - on the review tab, `Elaborate` preview expands to all question notes and all option notes, including notes on unselected options
 - transcript-friendly call and result rendering
+- `/answer` command to extract a raw-JSON `AskParams` form from the latest completed assistant message and open the ask UI
+- `/answer` extraction may use an internal `freeform: true` option for open-ended questions with no explicit choices; these render as user-input-only questions with the label `Type your answer:`, no numbered option row, and no selection caret; this marker is not part of the public `ask_user` tool contract
+- `/answer:again` command to replay the latest `/answer`-extracted form on the current branch
+- `/ask:replay` command to replay the latest real `ask_user` form on the current branch
 - ask settings list with binary behaviour toggles
 - `?` in the ask flow and `/ask-settings` in pi open the same lightweight ask settings overlay
 - behaviour settings persist immediately when changed: `Auto-submit when answered without notes`, `Confirm dismiss when dirty`, `Double-press review shortcuts`, and `Show footer hints`
@@ -241,6 +245,16 @@ Validation is handled inside the tool so malformed calls produce the same struct
 
 The ask flow subscribes to runtime settings updates while open. In practice, this means changing `Auto-submit when answered without notes`, `Confirm dismiss when dirty`, `Double-press review shortcuts`, `Show footer hints`, or reloading config-backed keymaps can affect the in-progress ask flow immediately instead of only future asks.
 
+## Slash command replay/extraction
+
+- valid `ask_user` payloads are persisted as branch custom entries before the UI opens, so `/ask:replay` can reopen them after cancel, `/resume`, or `/tree`
+- `/answer` scans the current branch for the latest assistant message; if that message did not finish with `stop`, extraction is refused
+- `/answer` expects the extractor to return raw JSON only; JSON parse failures are retried according to `answer.extractionRetries`, then reported to the user without opening the ask UI
+- `{ "questions": [] }` from extraction means no questions were found and is not treated as an invalid ask payload
+- command-flow cancellation closes with a notification and does not send a message to the agent
+- submitted or elaborated command-flow results are sent back with user-message semantics
+- replay commands scan only `ctx.sessionManager.getBranch()`, ignore sibling/future branch payloads, and revalidate stored payloads before opening the UI
+
 The fallback message includes normalized pending questions and options so the caller can re-ask them manually. `details.questions` still contains normalized question metadata, while `details.answers` stays empty until a user responds.
 
 ## Skill alignment (advisory)

package.json

@@ -84,7 +84,8 @@
   "peerDependencies": {
     "@mariozechner/pi-coding-agent": "*",
     "@mariozechner/pi-tui": "*",
-    "typebox": "*"
+    "typebox": "*",
+    "@mariozechner/pi-ai": "*"
   },
   "devDependencies": {
     "@biomejs/biome": "2.4.12",
@@ -105,7 +106,8 @@
     "lefthook": "^2.1.6",
     "semantic-release": "^25.0.3",
     "typescript": "6.0.2",
-    "ultracite": "7.6.0"
+    "ultracite": "7.6.0",
+    "@mariozechner/pi-ai": "0.69.0"
   },
   "config": {
     "commitizen": {

pnpm-lock.yaml

@@ -23,6 +23,9 @@ importers:
       '@commitlint/config-conventional':
         specifier: ^19.8.1
         version: 19.8.1
+      '@mariozechner/pi-ai':
+        specifier: 0.69.0
+        version: 0.69.0(ws@8.20.0)(zod@4.3.6)
       '@mariozechner/pi-coding-agent':
         specifier: 0.69.0
         version: 0.69.0(ws@8.20.0)(zod@4.3.6)

skills/ask-user/SKILL.md

@@ -87,8 +87,9 @@ After attempt 2:
 ## ask_user payload quality
 
 - Ask one concrete decision at a time.
-- Prefer 2-4 clear options (single-select by default).
+- Provide clear, distinct options. Do not add filler options.
-- Use `multi` only when selections are truly independent.
+- Choose question type from semantics: `single` means one answer is expected, `multi` means multiple answers could reasonably be selected, and `preview` means options need preview-pane detail with non-empty preview text.
+- Avoid defaulting mechanically; infer from whether options are mutually exclusive, can coexist, or need preview-pane detail.
 - Keep option labels short and outcome-oriented.
 - Include trade-off descriptions when non-obvious.
 - For research/planning, ask about goals, constraints, evaluation criteria, audience, budget, timeline, risk tolerance, and desired output only when they materially affect the result.

src/answer-commands.ts

@@ -0,0 +1,321 @@
+import {
+	BorderedLoader,
+	type ExtensionAPI,
+	type ExtensionCommandContext,
+	type ExtensionContext,
+} from "@mariozechner/pi-coding-agent";
+import {
+	extractAskParams,
+	selectExtractionModel,
+} from "./answer-extraction.ts";
+import {
+	type AskPayloadSource,
+	appendAskPayload,
+	findLatestPayloadInCurrentBranch,
+} from "./ask-payload-store.ts";
+import {
+	invalidPayloadResponse,
+	successfulResponse,
+	validateParams,
+} from "./ask-tool-helpers.ts";
+import { getAskConfigStore } from "./config/store.ts";
+import type { AskParams, AskResult } from "./types.ts";
+import { runAskFlow } from "./ui/controller.ts";
+
+interface AssistantTextSource {
+	entryId: string;
+	text: string;
+}
+
+type ExtractionSelection = Exclude<
+	Awaited<ReturnType<typeof selectExtractionModel>>,
+	{ error: string }
+>;
+
+type ExtractionUiResult =
+	| { cancelled: true }
+	| { error: string }
+	| { params: AskParams };
+
+export function registerAnswerCommands(pi: ExtensionAPI): void {
+	pi.registerCommand("answer", {
+		description:
+			"Extract questions from the latest assistant message into an ask form",
+		handler: async (_args, ctx) => runAnswerCommand(pi, ctx),
+	});
+
+	pi.registerCommand("answer:again", {
+		description: "Reopen the previous /answer form on this branch",
+		handler: async (_args, ctx) =>
+			runReplayCommand(pi, ctx, {
+				missingMessage:
+					"No previous /answer form found on this branch; use /answer first.",
+				noticePrefix: "Reopening previous /answer form on this branch",
+				source: "answer-extraction",
+			}),
+	});
+
+	pi.registerCommand("ask:replay", {
+		description: "Replay the previous ask_user form on this branch",
+		handler: async (_args, ctx) =>
+			runReplayCommand(pi, ctx, {
+				missingMessage: "No previous ask_user form found on this branch.",
+				noticePrefix: "Replaying previous ask_user form on this branch",
+				source: "tool",
+			}),
+	});
+}
+
+async function runAnswerCommand(
+	pi: ExtensionAPI,
+	ctx: ExtensionCommandContext
+): Promise<void> {
+	if (!ctx.hasUI) {
+		ctx.ui.notify("/answer requires interactive UI.", "error");
+		return;
+	}
+
+	const assistant = findLatestAssistantText(ctx);
+	if ("error" in assistant) {
+		ctx.ui.notify(assistant.error, "error");
+		return;
+	}
+
+	const config = await getAskConfigStore().getConfig();
+	const selected = await selectExtractionModel(
+		ctx,
+		config.answer.extractionModels
+	);
+	if ("error" in selected) {
+		ctx.ui.notify(selected.error, "error");
+		return;
+	}
+	if (selected.usedFallback) {
+		ctx.ui.notify(
+			`Configured extraction models unavailable; using current chat model ${selected.model.provider}/${selected.model.id}.`,
+			"info"
+		);
+	}
+
+	const params = await extractAndValidateAnswerParams(
+		ctx,
+		assistant,
+		selected,
+		config
+	);
+	if (!params) {
+		return;
+	}
+
+	appendAskPayload(pi, {
+		params,
+		source: "answer-extraction",
+		sourceEntryId: assistant.entryId,
+	});
+
+	await runAskAndSendSubmittedResult(pi, ctx, params, {
+		allowFreeform: true,
+	});
+}
+
+async function extractAndValidateAnswerParams(
+	ctx: ExtensionCommandContext,
+	assistant: AssistantTextSource,
+	selected: ExtractionSelection,
+	config: Awaited<ReturnType<ReturnType<typeof getAskConfigStore>["getConfig"]>>
+): Promise<AskParams | undefined> {
+	const extraction = await runExtractionUi(ctx, assistant, selected, config);
+	if ("cancelled" in extraction) {
+		ctx.ui.notify("Question extraction cancelled.", "info");
+		return;
+	}
+	if ("error" in extraction) {
+		ctx.ui.notify(formatExtractionError(extraction.error), "error");
+		return;
+	}
+	if (!isAskParamsCandidate(extraction.params)) {
+		ctx.ui.notify("Question extraction returned an invalid ask form.", "error");
+		return;
+	}
+	if (extraction.params.questions.length === 0) {
+		ctx.ui.notify(
+			"No questions found in the latest assistant message.",
+			"info"
+		);
+		return;
+	}
+	const validation = validateParams(extraction.params, { allowFreeform: true });
+	if (!validation.ok) {
+		const response = invalidPayloadResponse(
+			extraction.params,
+			validation.issues
+		);
+		ctx.ui.notify(response.content[0].text, "error");
+		return;
+	}
+	return extraction.params;
+}
+
+function runExtractionUi(
+	ctx: ExtensionCommandContext,
+	assistant: AssistantTextSource,
+	selected: ExtractionSelection,
+	config: Awaited<ReturnType<ReturnType<typeof getAskConfigStore>["getConfig"]>>
+): Promise<ExtractionUiResult> {
+	return ctx.ui.custom<ExtractionUiResult>((tui, theme, _keybindings, done) => {
+		let completed = false;
+		const doneOnce = (result: ExtractionUiResult) => {
+			if (completed) {
+				return;
+			}
+			completed = true;
+			done(result);
+		};
+		const loader = new BorderedLoader(
+			tui,
+			theme,
+			`Extracting questions using ${selected.model.provider}/${selected.model.id}...`
+		);
+		loader.onAbort = () => doneOnce({ cancelled: true });
+		extractAskParams({
+			assistantText: assistant.text,
+			auth: selected.auth,
+			model: selected.model,
+			onRetry: (attempt, maxRetries) => {
+				ctx.ui.notify(
+					`Retrying extraction JSON repair (${attempt}/${maxRetries})...`,
+					"info"
+				);
+			},
+			retries: config.answer.extractionRetries,
+			signal: loader.signal,
+			timeoutMs: config.answer.extractionTimeoutMs,
+		})
+			.then((params) => doneOnce({ params }))
+			.catch((error) => doneOnce({ error: formatErrorMessage(error) }));
+		return loader;
+	});
+}
+
+function formatErrorMessage(error: unknown): string {
+	if (error instanceof Error) {
+		return error.name ? `${error.name}: ${error.message}` : error.message;
+	}
+	return String(error);
+}
+
+function isAskParamsCandidate(value: unknown): value is AskParams {
+	return (
+		!!value &&
+		typeof value === "object" &&
+		Array.isArray((value as { questions?: unknown }).questions)
+	);
+}
+
+function formatExtractionError(error: string): string {
+	if (error.includes("valid JSON")) {
+		return error;
+	}
+	if (error.includes("cancelled")) {
+		return "Question extraction cancelled.";
+	}
+	if (
+		error.includes("aborted") ||
+		error.includes("AbortError") ||
+		error.includes("timed out")
+	) {
+		return "Question extraction timed out. Try again or configure a faster extraction model.";
+	}
+	return error;
+}
+
+async function runReplayCommand(
+	pi: ExtensionAPI,
+	ctx: ExtensionCommandContext,
+	options: {
+		missingMessage: string;
+		noticePrefix: string;
+		source: AskPayloadSource;
+	}
+): Promise<void> {
+	if (!ctx.hasUI) {
+		ctx.ui.notify("Ask replay requires interactive UI.", "error");
+		return;
+	}
+	const lookup = findLatestPayloadInCurrentBranch(ctx, options.source);
+	if (!lookup.data) {
+		ctx.ui.notify(
+			lookup.invalidMatchFound
+				? "Previous form exists on this branch but is no longer compatible."
+				: options.missingMessage,
+			"info"
+		);
+		return;
+	}
+	ctx.ui.notify(
+		`${options.noticePrefix}: ${lookup.data.params.questions.length} question(s).`,
+		"info"
+	);
+	await runAskAndSendSubmittedResult(pi, ctx, lookup.data.params, {
+		allowFreeform: options.source === "answer-extraction",
+	});
+}
+
+async function runAskAndSendSubmittedResult(
+	pi: Pick<ExtensionAPI, "sendUserMessage">,
+	ctx: ExtensionContext,
+	params: AskParams,
+	options: { allowFreeform: boolean }
+): Promise<void> {
+	const result = await runAskFlow(ctx, params, options);
+	if (result.cancelled) {
+		ctx.ui.notify("Ask form cancelled.", "info");
+		return;
+	}
+	sendAskResult(pi, result, ctx);
+}
+
+function sendAskResult(
+	pi: Pick<ExtensionAPI, "sendUserMessage">,
+	result: AskResult,
+	ctx: Pick<ExtensionContext, "isIdle">
+): void {
+	const response = successfulResponse(result);
+	const text = response.content[0];
+	pi.sendUserMessage(
+		text.text,
+		ctx.isIdle() ? undefined : { deliverAs: "followUp" }
+	);
+}
+
+function findLatestAssistantText(
+	ctx: Pick<ExtensionContext, "sessionManager">
+): AssistantTextSource | { error: string } {
+	const branch = ctx.sessionManager.getBranch();
+	for (let index = branch.length - 1; index >= 0; index--) {
+		const entry = branch[index];
+		if (entry.type !== "message") {
+			continue;
+		}
+		const message = entry.message;
+		if (!("role" in message) || message.role !== "assistant") {
+			continue;
+		}
+		if (message.stopReason !== "stop") {
+			return {
+				error: `Latest assistant message is incomplete (${message.stopReason}); wait for it to finish, then run /answer again.`,
+			};
+		}
+		const text = message.content
+			.filter(
+				(part): part is { text: string; type: "text" } => part.type === "text"
+			)
+			.map((part) => part.text)
+			.join("\n")
+			.trim();
+		if (text) {
+			return { entryId: entry.id, text };
+		}
+	}
+	return { error: "No assistant message found to extract questions from." };
+}

src/answer-extraction.ts

@@ -0,0 +1,354 @@
+import {
+	type Api,
+	complete,
+	type Model,
+	type UserMessage,
+} from "@mariozechner/pi-ai";
+import type { ExtensionContext } from "@mariozechner/pi-coding-agent";
+import type { AskConfig } from "./config/schema.ts";
+import type { AskParams } from "./types.ts";
+
+export const ANSWER_EXTRACTION_SYSTEM_PROMPT = `You extract user-input questions from an assistant message and return ONLY raw JSON.
+
+Return JSON matching this TypeScript shape:
+{
+  "title"?: string,
+  "questions": [
+    {
+      "id": string,
+      "label"?: string,
+      "prompt": string,
+      "type"?: "single" | "multi" | "preview",
+      "required"?: boolean,
+      "options": [
+        {
+          "value": string,
+          "label": string,
+          "description"?: string,
+          "preview"?: string,
+          "freeform"?: boolean
+        }
+      ]
+    }
+  ]
+}
+
+Rules:
+- Output raw JSON only. No Markdown. No prose. No code fences.
+- Extract questions that require user input.
+- Ignore generic conversational or clarification prompts such as "How can I help?", "Could you clarify?", or "Let me know what you need" unless they include concrete choices.
+- Extract a question when it has explicit choices or when the user should type their own answer.
+- Preserve question order.
+- Generate stable snake_case ids.
+- Choose question type from the question semantics.
+- Use type "single" when one answer is expected.
+- Use type "multi" when multiple answers could reasonably be selected.
+- Use type "preview" when options need preview-pane detail and every option has non-empty preview text.
+- Avoid defaulting mechanically; infer from whether the options are mutually exclusive, can coexist, or need preview-pane detail.
+- Each extracted question must have at least one option.
+- Options rule:
+  - Extract options only from choices explicitly offered by the assistant.
+  - If the assistant gives concrete choices, use those choices as normal options.
+  - If the assistant gives examples only, do not treat examples as choices unless they are presented as selectable answers.
+  - If no concrete choices are given and the user should type their own answer, create exactly one freeform option: {"value":"freeform","label":"Type answer","freeform":true}.
+  - Never invent options.
+  - Never mix a freeform option with normal options.
+- Provide clear, distinct options. Do not add filler options.
+- Do not create an option that merely restates the question.
+- Include descriptions only when helpful.
+- Return {"questions":[]} if no questions are found.`;
+
+interface SelectedExtractionModel {
+	auth: { apiKey?: string; headers?: Record<string, string> };
+	model: Model<Api>;
+	usedFallback: boolean;
+}
+
+export async function selectExtractionModel(
+	ctx: Pick<ExtensionContext, "model" | "modelRegistry">,
+	preferences: AskConfig["answer"]["extractionModels"]
+): Promise<SelectedExtractionModel | { error: string }> {
+	for (const preference of preferences) {
+		const model = ctx.modelRegistry.find(preference.provider, preference.id);
+		if (!model) {
+			continue;
+		}
+		const auth = await ctx.modelRegistry.getApiKeyAndHeaders(model);
+		if (auth.ok) {
+			return { model, auth, usedFallback: false };
+		}
+	}
+
+	if (!ctx.model) {
+		return {
+			error:
+				"No available extraction model. Configure answer.extractionModels or select a chat model.",
+		};
+	}
+
+	const auth = await ctx.modelRegistry.getApiKeyAndHeaders(ctx.model);
+	if (!auth.ok) {
+		return {
+			error: `No auth for fallback chat model: ${ctx.model.provider}/${ctx.model.id}.`,
+		};
+	}
+
+	return { model: ctx.model, auth, usedFallback: true };
+}
+
+export async function extractAskParams(options: {
+	assistantText: string;
+	model: Model<Api>;
+	auth: { apiKey?: string; headers?: Record<string, string> };
+	retries: number;
+	signal?: AbortSignal;
+	timeoutMs: number;
+	onRetry?: (attempt: number, maxRetries: number) => void;
+}): Promise<AskParams> {
+	let lastCandidate: AskParams | undefined;
+	let lastResponse = "";
+	let lastError = "";
+	for (let attempt = 0; attempt <= options.retries; attempt++) {
+		if (attempt > 0) {
+			options.onRetry?.(attempt, options.retries);
+		}
+		const responseText = await runExtractionAttempt({
+			...options,
+			attempt,
+			lastError,
+			lastResponse,
+		});
+		lastResponse = responseText;
+		const parsed = parseExtractionCandidate(responseText);
+		if (!parsed.ok) {
+			lastError = parsed.error;
+			continue;
+		}
+		lastCandidate = parsed.params;
+		if (parsed.issues.length === 0) {
+			return parsed.params;
+		}
+		lastError = parsed.issues.join("\n");
+	}
+	if (lastCandidate) {
+		return repairExtractionParams(lastCandidate);
+	}
+	throw new Error(
+		"Question extraction did not return valid JSON after retries."
+	);
+}
+
+async function runExtractionAttempt(options: {
+	assistantText: string;
+	attempt: number;
+	auth: { apiKey?: string; headers?: Record<string, string> };
+	lastError: string;
+	lastResponse: string;
+	model: Model<Api>;
+	signal?: AbortSignal;
+	timeoutMs: number;
+}): Promise<string> {
+	const controller = new AbortController();
+	let timedOut = false;
+	const timeout = setTimeout(() => {
+		timedOut = true;
+		controller.abort();
+	}, options.timeoutMs);
+	const abortFromParent = () => controller.abort();
+	options.signal?.addEventListener("abort", abortFromParent, { once: true });
+	try {
+		const userMessage: UserMessage = {
+			role: "user",
+			content: [
+				{
+					type: "text",
+					text:
+						options.attempt === 0
+							? options.assistantText
+							: formatRetryPrompt(options),
+				},
+			],
+			timestamp: Date.now(),
+		};
+		const response = await complete(
+			options.model,
+			{
+				systemPrompt: ANSWER_EXTRACTION_SYSTEM_PROMPT,
+				messages: [userMessage],
+			},
+			{
+				apiKey: options.auth.apiKey,
+				headers: options.auth.headers,
+				signal: controller.signal,
+			}
+		);
+		if (response.stopReason === "aborted") {
+			throw new Error(
+				timedOut
+					? "Question extraction timed out. Try again or configure a faster extraction model."
+					: "Question extraction cancelled."
+			);
+		}
+		return response.content
+			.filter(
+				(part): part is { text: string; type: "text" } => part.type === "text"
+			)
+			.map((part) => part.text)
+			.join("\n");
+	} finally {
+		clearTimeout(timeout);
+		options.signal?.removeEventListener("abort", abortFromParent);
+	}
+}
+
+const MAX_EXTRACTED_OPTIONS_PER_QUESTION = 4;
+
+function parseExtractionCandidate(
+	responseText: string
+):
+	| { ok: true; params: AskParams; issues: string[] }
+	| { ok: false; error: string } {
+	try {
+		const parsed = JSON.parse(responseText.trim()) as unknown;
+		if (!isAskParamsLike(parsed)) {
+			return { ok: false, error: "root.questions must be an array" };
+		}
+		return {
+			ok: true,
+			params: parsed,
+			issues: collectExtractionBusinessIssues(parsed),
+		};
+	} catch (error) {
+		return {
+			ok: false,
+			error: error instanceof Error ? error.message : String(error),
+		};
+	}
+}
+
+function isAskParamsLike(value: unknown): value is AskParams {
+	return (
+		!!value &&
+		typeof value === "object" &&
+		Array.isArray((value as { questions?: unknown }).questions)
+	);
+}
+
+export function collectExtractionBusinessIssues(params: AskParams): string[] {
+	const issues: string[] = [];
+	params.questions.forEach((question, questionIndex) => {
+		if (isGenericConversationalPrompt(question.prompt)) {
+			issues.push(
+				`questions[${questionIndex}].prompt is generic conversational text; omit this question unless it includes concrete choices`
+			);
+		}
+		if (question.options.length > MAX_EXTRACTED_OPTIONS_PER_QUESTION) {
+			issues.push(
+				`questions[${questionIndex}].options has ${question.options.length} items; max is ${MAX_EXTRACTED_OPTIONS_PER_QUESTION}`
+			);
+		}
+		if (
+			question.options.length === 1 &&
+			optionRestatesQuestion(question.options[0]?.label, question.prompt)
+		) {
+			issues.push(
+				`questions[${questionIndex}].options[0] merely restates the question; omit this question or provide meaningful distinct options`
+			);
+		}
+	});
+	return issues;
+}
+
+export function repairExtractionParams(params: AskParams): AskParams {
+	return {
+		...params,
+		questions: params.questions
+			.filter((question) => !isGenericConversationalPrompt(question.prompt))
+			.filter(
+				(question) =>
+					question.options.length !== 1 ||
+					!optionRestatesQuestion(question.options[0]?.label, question.prompt)
+			)
+			.map((question) => ({
+				...question,
+				options: capOptionsPreservingOther(question.options),
+			})),
+	};
+}
+
+function capOptionsPreservingOther<T extends { label: string; value: string }>(
+	options: T[]
+): T[] {
+	if (options.length <= MAX_EXTRACTED_OPTIONS_PER_QUESTION) {
+		return options;
+	}
+	const other = options.find((option) => isOtherOption(option));
+	if (!other) {
+		return options.slice(0, MAX_EXTRACTED_OPTIONS_PER_QUESTION);
+	}
+	const head = options
+		.filter((option) => option !== other)
+		.slice(0, MAX_EXTRACTED_OPTIONS_PER_QUESTION - 1);
+	return [...head, other];
+}
+
+function isOtherOption(option: { label: string; value: string }): boolean {
+	const label = normalizeText(option.label);
+	const value = normalizeText(option.value);
+	return (
+		label === "other" || label.includes("something else") || value === "other"
+	);
+}
+
+function optionRestatesQuestion(
+	label: string | undefined,
+	prompt: string
+): boolean {
+	if (!label) {
+		return false;
+	}
+	const normalizedLabel = normalizeText(label);
+	const normalizedPrompt = normalizeText(prompt);
+	return (
+		normalizedLabel.length > 8 &&
+		(normalizedPrompt.includes(normalizedLabel) ||
+			normalizedLabel.includes(normalizedPrompt))
+	);
+}
+
+function isGenericConversationalPrompt(prompt: string): boolean {
+	const normalized = normalizeText(prompt);
+	return [
+		"how can i help",
+		"could you clarify",
+		"let me know what you need",
+		"what else is on your mind",
+		"anything on your mind i can help with",
+	].some((phrase) => normalized.includes(phrase));
+}
+
+function normalizeText(value: string): string {
+	return value
+		.toLowerCase()
+		.replace(/[^a-z0-9]+/g, " ")
+		.trim();
+}
+
+function formatRetryPrompt(options: {
+	assistantText: string;
+	lastError: string;
+	lastResponse: string;
+}): string {
+	return `Your previous response was not valid raw JSON.
+
+JSON.parse error:
+${options.lastError}
+
+Previous response:
+${options.lastResponse}
+
+Original assistant message:
+${options.assistantText}
+
+Return ONLY valid raw JSON matching the schema and fix the reported issues. No Markdown. No prose.`;
+}

src/ask-payload-store.ts

@@ -0,0 +1,86 @@
+import type {
+	ExtensionAPI,
+	ExtensionContext,
+} from "@mariozechner/pi-coding-agent";
+import { validateParams } from "./ask-tool-helpers.ts";
+import type { AskParams } from "./types.ts";
+
+export const ASK_PAYLOAD_ENTRY_TYPE = "ask:payload";
+export const ASK_PAYLOAD_ENTRY_VERSION = 1;
+
+export type AskPayloadSource = "answer-extraction" | "tool";
+
+export interface AskPayloadEntryData {
+	params: AskParams;
+	source: AskPayloadSource;
+	sourceEntryId?: string;
+	timestamp: number;
+	version: typeof ASK_PAYLOAD_ENTRY_VERSION;
+}
+
+export function appendAskPayload(
+	pi: Pick<ExtensionAPI, "appendEntry">,
+	data: Omit<AskPayloadEntryData, "timestamp" | "version">
+): void {
+	pi.appendEntry(ASK_PAYLOAD_ENTRY_TYPE, {
+		version: ASK_PAYLOAD_ENTRY_VERSION,
+		timestamp: Date.now(),
+		...data,
+	});
+}
+
+export function findLatestPayloadInCurrentBranch(
+	ctx: Pick<ExtensionContext, "sessionManager">,
+	source: AskPayloadSource
+): { data?: AskPayloadEntryData; invalidMatchFound: boolean } {
+	let invalidMatchFound = false;
+	for (const entry of [...ctx.sessionManager.getBranch()].reverse()) {
+		if (!isAskPayloadEntry(entry)) {
+			continue;
+		}
+		const data = entry.data;
+		if (data?.source !== source) {
+			continue;
+		}
+		if (isValidAskPayloadData(data)) {
+			return { data, invalidMatchFound };
+		}
+		invalidMatchFound = true;
+	}
+	return { invalidMatchFound };
+}
+
+function isAskPayloadEntry(entry: unknown): entry is {
+	customType: string;
+	data?: Partial<AskPayloadEntryData>;
+	type: "custom";
+} {
+	return (
+		!!entry &&
+		typeof entry === "object" &&
+		(entry as { type?: unknown }).type === "custom" &&
+		(entry as { customType?: unknown }).customType === ASK_PAYLOAD_ENTRY_TYPE
+	);
+}
+
+function isValidAskPayloadData(data: unknown): data is AskPayloadEntryData {
+	if (!(data && typeof data === "object")) {
+		return false;
+	}
+	const payload = data as Partial<AskPayloadEntryData>;
+	if (payload.version !== ASK_PAYLOAD_ENTRY_VERSION) {
+		return false;
+	}
+	if (payload.source !== "tool" && payload.source !== "answer-extraction") {
+		return false;
+	}
+	if (
+		!payload.params ||
+		validateParams(payload.params, {
+			allowFreeform: payload.source === "answer-extraction",
+		}).ok === false
+	) {
+		return false;
+	}
+	return true;
+}

src/ask-tool-helpers.ts

@@ -18,28 +18,34 @@ export const ASK_TOOL_DESCRIPTION =
 export const ASK_TOOL_PROMPT_GUIDELINES = [
 	"Use this tool before making preference-sensitive decisions about scope, tone, UX, naming, architecture, docs, or implementation direction.",
 	"When multiple valid directions exist, ask 1-3 concise questions instead of committing to one path on your own.",
-	"Prefer one focused decision per question and use short labels with 2-4 clear options.",
+	"Prefer one focused decision per question. Use short labels. Provide clear, distinct options. Do not add filler options.",
 	"Always include a non-empty `value` for every option.",
-	'Use `type: "single"` by default and `type: "multi"` only when several answers can genuinely apply.',
+	"Choose question `type` from the question semantics: `single` means one answer is expected, `multi` means multiple answers could reasonably be selected, and `preview` means options need preview-pane detail and every option includes non-empty `preview` text.",
-	'Use `type: "preview"` only when every option includes non-empty `preview` text for the dedicated preview pane. Option descriptions do not satisfy this requirement.',
+	"Avoid defaulting mechanically; infer from whether the options are mutually exclusive, can coexist, or need preview-pane detail.",
+	'Use `type: "preview"` only when every option includes non-empty `preview` text. Option descriptions do not satisfy this requirement.',
 	"After clarifying a note or follow-up question, prefer another structured ask_user follow-up if a choice is still needed instead of switching to plain-text multiple choice in chat.",
 	"When prior answers already narrow the branch, bundle the next 2-3 related unresolved decisions into one follow-up ask instead of issuing a long sequence of single-question asks.",
 	"Use one-at-a-time follow-up asks only when the next question materially depends on the previous answer.",
 ] as const;
 
+interface ValidateParamsOptions {
+	allowFreeform?: boolean;
+}
+
 export function validateParams(
-	params: AskParams
+	params: AskParams,
+	options: ValidateParamsOptions = {}
 ):
 	| { ok: true; state: ReturnType<typeof createInitialState> }
 	| { ok: false; issues: AskValidationIssue[] } {
-	const issues = collectValidationIssues(params);
+	const issues = collectValidationIssues(params, options);
 	if (issues.length > 0) {
 		return { ok: false, issues };
 	}
 
 	return {
 		ok: true,
-		state: createInitialState(params),
+		state: createInitialState(params, options),
 	};
 }
 

src/ask-tool.ts

@@ -2,6 +2,7 @@ import type {
 	ExtensionAPI,
 	ExtensionContext,
 } from "@mariozechner/pi-coding-agent";
+import { appendAskPayload } from "./ask-payload-store.ts";
 import {
 	ASK_TOOL_DESCRIPTION,
 	ASK_TOOL_PROMPT_GUIDELINES,
@@ -25,16 +26,25 @@ export function registerAskTool(pi: ExtensionAPI) {
 			"Clarify ambiguous or preference-sensitive decisions with a short interactive interview before proceeding",
 		promptGuidelines: [...ASK_TOOL_PROMPT_GUIDELINES],
 		parameters: AskParamsSchema,
-		execute: executeAskTool,
+		execute: (toolCallId, params, signal, onUpdate, ctx) =>
+			executeAskTool(
+				pi,
+				toolCallId,
+				params as AskParams,
+				signal,
+				onUpdate,
+				ctx
+			),
 		renderCall: renderAskToolCall,
 		renderResult: renderAskToolResult,
 	});
 }
 
-interface ExecuteContext extends Pick<ExtensionContext, "cwd" | "hasUI"> {}
+interface ExecuteContext extends ExtensionContext {}
 
 async function executeAskTool(
-	_toolCallId: string,
+	pi: Pick<ExtensionAPI, "appendEntry">,
+	toolCallId: string,
 	params: AskParams,
 	_signal: AbortSignal | undefined,
 	_onUpdate: unknown,
@@ -44,6 +54,11 @@ async function executeAskTool(
 	if (!validation.ok) {
 		return invalidPayloadResponse(params, validation.issues);
 	}
+	appendAskPayload(pi, {
+		params,
+		source: "tool",
+		sourceEntryId: toolCallId,
+	});
 	if (!ctx.hasUI) {
 		return nonInteractiveResponse(validation.state);
 	}

src/config/defaults.ts

@@ -1,7 +1,20 @@
 import { DEFAULT_ASK_KEYMAPS } from "../constants/keymaps.ts";
-import type { AskConfig, AskConfigFileV1 } from "./schema.ts";
+import type { AskConfig, AskConfigFileV2 } from "./schema.ts";
+
+const DEFAULT_EXTRACTION_RETRIES = 1;
+const MAX_EXTRACTION_RETRIES = 3;
+const DEFAULT_EXTRACTION_TIMEOUT_MS = 30_000;
 
 export const DEFAULT_ASK_CONFIG: AskConfig = {
+	answer: {
+		extractionModels: [
+			{ provider: "openai-codex", id: "gpt-5.4-mini" },
+			{ provider: "github-copilot", id: "gpt-5.4-mini" },
+			{ provider: "anthropic", id: "claude-haiku-4-5" },
+		],
+		extractionRetries: DEFAULT_EXTRACTION_RETRIES,
+		extractionTimeoutMs: DEFAULT_EXTRACTION_TIMEOUT_MS,
+	},
 	behaviour: {
 		autoSubmitWhenAnsweredWithoutNotes: false,
 		confirmDismissWhenDirty: true,
@@ -14,9 +27,24 @@ export const DEFAULT_ASK_CONFIG: AskConfig = {
 };
 
 export function normalizeAskConfig(
-	config?: Partial<AskConfigFileV1> | AskConfig
+	config?: Partial<AskConfigFileV2> | AskConfig
 ): AskConfig {
 	return {
+		answer: {
+			extractionModels:
+				config?.answer?.extractionModels?.filter(isValidModelPreference) ??
+				DEFAULT_ASK_CONFIG.answer.extractionModels,
+			extractionRetries: clampInteger(
+				config?.answer?.extractionRetries,
+				0,
+				MAX_EXTRACTION_RETRIES,
+				DEFAULT_EXTRACTION_RETRIES
+			),
+			extractionTimeoutMs: positiveNumberOrDefault(
+				config?.answer?.extractionTimeoutMs,
+				DEFAULT_EXTRACTION_TIMEOUT_MS
+			),
+		},
 		behaviour: {
 			autoSubmitWhenAnsweredWithoutNotes:
 				config?.behaviour?.autoSubmitWhenAnsweredWithoutNotes ??
@@ -38,10 +66,15 @@ export function normalizeAskConfig(
 	};
 }
 
-export function toAskConfigFileV1(config: AskConfig): AskConfigFileV1 {
+export function toAskConfigFileV2(config: AskConfig): AskConfigFileV2 {
 	const normalized = normalizeAskConfig(config);
 	return {
-		schemaVersion: 1,
+		schemaVersion: 2,
+		answer: {
+			extractionModels: normalized.answer.extractionModels,
+			extractionRetries: normalized.answer.extractionRetries,
+			extractionTimeoutMs: normalized.answer.extractionTimeoutMs,
+		},
 		behaviour: {
 			autoSubmitWhenAnsweredWithoutNotes:
 				normalized.behaviour.autoSubmitWhenAnsweredWithoutNotes,
@@ -60,3 +93,35 @@ export function toAskConfigFileV1(config: AskConfig): AskConfigFileV1 {
 		},
 	};
 }
+
+function isValidModelPreference(value: unknown): value is {
+	id: string;
+	provider: string;
+} {
+	return (
+		!!value &&
+		typeof value === "object" &&
+		typeof (value as { id?: unknown }).id === "string" &&
+		(value as { id: string }).id.trim().length > 0 &&
+		typeof (value as { provider?: unknown }).provider === "string" &&
+		(value as { provider: string }).provider.trim().length > 0
+	);
+}
+
+function clampInteger(
+	value: unknown,
+	min: number,
+	max: number,
+	fallback: number
+): number {
+	if (typeof value !== "number" || !Number.isFinite(value)) {
+		return fallback;
+	}
+	return Math.max(min, Math.min(max, Math.trunc(value)));
+}
+
+function positiveNumberOrDefault(value: unknown, fallback: number): number {
+	return typeof value === "number" && Number.isFinite(value) && value > 0
+		? value
+		: fallback;
+}

src/config/migrate.ts

@@ -8,8 +8,8 @@ import {
 	type AskConfigVersionMigrationResult,
 	migrateAskConfigFileToCurrent,
 } from "./migrations/index.ts";
-import type { AskConfig, AskConfigFileV1 } from "./schema.ts";
+import type { AskConfig, AskConfigFileV2 } from "./schema.ts";
-import { validateAskConfigFileV1 } from "./schema.ts";
+import { validateAskConfigFileV2 } from "./schema.ts";
 
 export class AskConfigMigrationError extends Error {
 	readonly reason: "invalid_or_unsupported" | "migration_failed";
@@ -42,14 +42,14 @@ export function migrateAskConfig(raw: unknown): AskConfigMigrationResult {
 		);
 	}
 
-	if (!validateAskConfigFileV1.Check(migratedFile.config)) {
+	if (!validateAskConfigFileV2.Check(migratedFile.config)) {
 		throw new AskConfigMigrationError(
 			"Config was invalid or unsupported.",
 			"invalid_or_unsupported"
 		);
 	}
 
-	const currentFile = migratedFile.config as AskConfigFileV1;
+	const currentFile = migratedFile.config as AskConfigFileV2;
 	const config = normalizeAskConfig(currentFile);
 	const keymapsResult = normalizeConfiguredKeymaps(
 		currentFile.keymaps as

src/config/migrations/index.ts

@@ -1,8 +1,17 @@
 import type { AskConfigMigration, VersionedAskConfigFile } from "./types.ts";
 
-export const CURRENT_ASK_CONFIG_SCHEMA_VERSION = 1;
+export const CURRENT_ASK_CONFIG_SCHEMA_VERSION = 2;
 
-const ASK_CONFIG_MIGRATIONS: AskConfigMigration[] = [];
+const ASK_CONFIG_MIGRATIONS: AskConfigMigration[] = [
+	{
+		from: 1,
+		to: 2,
+		migrate: (config) => ({
+			...config,
+			schemaVersion: 2,
+		}),
+	},
+];
 
 export class AskConfigVersionMigrationError extends Error {
 	readonly reason: "invalid_or_unsupported" | "migration_failed";

src/config/schema.ts

@@ -1,8 +1,22 @@
 import { type Static, Type } from "typebox";
 import { Compile } from "typebox/compile";
 
-export const AskConfigFileV1Schema = Type.Object({
+const AskAnswerModelPreferenceSchema = Type.Object({
-	schemaVersion: Type.Literal(1),
+	id: Type.String(),
+	provider: Type.String(),
+});
+
+export const AskConfigFileV2Schema = Type.Object({
+	schemaVersion: Type.Literal(2),
+	answer: Type.Optional(
+		Type.Object({
+			extractionModels: Type.Optional(
+				Type.Array(AskAnswerModelPreferenceSchema)
+			),
+			extractionRetries: Type.Optional(Type.Number()),
+			extractionTimeoutMs: Type.Optional(Type.Number()),
+		})
+	),
 	behaviour: Type.Optional(
 		Type.Object({
 			autoSubmitWhenAnsweredWithoutNotes: Type.Optional(Type.Boolean()),
@@ -14,16 +28,27 @@ export const AskConfigFileV1Schema = Type.Object({
 	keymaps: Type.Optional(
 		Type.Object({
 			cancel: Type.Optional(Type.String()),
-			dismiss: Type.Optional(Type.String()),
-			toggle: Type.Optional(Type.String()),
 			confirm: Type.Optional(Type.String()),
+			dismiss: Type.Optional(Type.String()),
 			optionNote: Type.Optional(Type.String()),
 			questionNote: Type.Optional(Type.String()),
+			toggle: Type.Optional(Type.String()),
 		})
 	),
 });
 
-export type AskConfigFileV1 = Static<typeof AskConfigFileV1Schema>;
+export type AskConfigFileV2 = Static<typeof AskConfigFileV2Schema>;
+export type AskConfigFileV1 = Omit<
+	AskConfigFileV2,
+	"answer" | "schemaVersion"
+> & {
+	schemaVersion: 1;
+};
+
+export interface AskAnswerModelPreference {
+	id: string;
+	provider: string;
+}
 
 export interface AskConfigKeymaps {
 	cancel: string;
@@ -35,6 +60,11 @@ export interface AskConfigKeymaps {
 }
 
 export interface AskConfig {
+	answer: {
+		extractionModels: AskAnswerModelPreference[];
+		extractionRetries: number;
+		extractionTimeoutMs: number;
+	};
 	behaviour: {
 		autoSubmitWhenAnsweredWithoutNotes: boolean;
 		confirmDismissWhenDirty: boolean;
@@ -44,4 +74,4 @@ export interface AskConfig {
 	keymaps: AskConfigKeymaps;
 }
 
-export const validateAskConfigFileV1 = Compile(AskConfigFileV1Schema);
+export const validateAskConfigFileV2 = Compile(AskConfigFileV2Schema);

src/config/store.ts

@@ -7,7 +7,7 @@ import { getAgentDir } from "@mariozechner/pi-coding-agent";
 import {
 	DEFAULT_ASK_CONFIG,
 	normalizeAskConfig,
-	toAskConfigFileV1,
+	toAskConfigFileV2,
 } from "./defaults.ts";
 import { AskConfigMigrationError, migrateAskConfig } from "./migrate.ts";
 import { migrateAskConfigPathIfNeeded } from "./path-migrations.ts";
@@ -63,10 +63,10 @@ export class AskConfigStore {
 		return (await this.ensureLoaded()).config;
 	}
 
-	async save(config: AskConfig): Promise<AskConfig> {
+	async save(config: AskConfig | Partial<AskConfig>): Promise<AskConfig> {
 		const normalized = normalizeAskConfig(config);
 		const content = JSON.stringify(
-			toAskConfigFileV1(normalized),
+			toAskConfigFileV2(normalized),
 			null,
 			2
 		).concat("\n");

src/index.ts

@@ -1,6 +1,7 @@
 import { dirname, resolve } from "node:path";
 import { fileURLToPath } from "node:url";
 import type { ExtensionAPI } from "@mariozechner/pi-coding-agent";
+import { registerAnswerCommands } from "./answer-commands.ts";
 import { registerAskSettingsCommand } from "./ask-settings-command.ts";
 import { registerAskTool } from "./ask-tool.ts";
 
@@ -18,4 +19,5 @@ export default function askExtension(pi: ExtensionAPI) {
 	}));
 	registerAskTool(pi);
 	registerAskSettingsCommand(pi);
+	registerAnswerCommands(pi);
 }

src/schema.ts

@@ -46,7 +46,7 @@ export const AskQuestionSchema = Type.Object({
 	type: Type.Optional(
 		Type.String({
 			description:
-				"Question type: use `single` by default. Use `preview` only when every option includes `preview` text for the dedicated preview pane; descriptions alone are not enough.",
+				"Question type: `single` means one answer is expected, `multi` means multiple answers could reasonably be selected, and `preview` means options need preview-pane detail. Use `preview` only when every option includes `preview` text; descriptions alone are not enough.",
 		})
 	),
 	required: Type.Optional(
@@ -56,7 +56,8 @@ export const AskQuestionSchema = Type.Object({
 		})
 	),
 	options: Type.Array(AskOptionSchema, {
-		description: "Answer options; prefer 2-4 clear choices when possible",
+		description:
+			"Answer options; provide clear, distinct choices and do not add filler options",
 	}),
 });
 

src/state/create.ts

@@ -2,9 +2,16 @@ import type { AskParams } from "../types.ts";
 import { normalizeQuestions } from "./normalize.ts";
 import { createInitialState as createBaseState } from "./transitions.ts";
 
-export function createInitialState(params: AskParams) {
+interface CreateInitialStateOptions {
+	allowFreeform?: boolean;
+}
+
+export function createInitialState(
+	params: AskParams,
+	options: CreateInitialStateOptions = {}
+) {
 	return createBaseState({
 		title: params.title,
-		questions: normalizeQuestions(params),
+		questions: normalizeQuestions(params, options),
 	});
 }

src/state/editor.ts

@@ -62,5 +62,9 @@ export function syncStateToSelection(state: AskState): AskState {
 		return state;
 	}
 
+	if (getAnswer(state, question.id)?.customText?.trim()) {
+		return state;
+	}
+
 	return enterInputMode(state, question.id);
 }

src/state/normalize.ts

@@ -11,8 +11,15 @@ interface IssueCollector {
 	issues: AskValidationIssue[];
 }
 
-export function normalizeQuestions(params: AskParams): AskQuestion[] {
+interface ValidationOptions {
-	const issues = collectValidationIssues(params);
+	allowFreeform?: boolean;
+}
+
+export function normalizeQuestions(
+	params: AskParams,
+	options: ValidationOptions = {}
+): AskQuestion[] {
+	const issues = collectValidationIssues(params, options);
 	if (issues.length > 0) {
 		throw new Error(issues[0]?.message ?? "Invalid ask_user payload");
 	}
@@ -20,10 +27,11 @@ export function normalizeQuestions(params: AskParams): AskQuestion[] {
 }
 
 export function collectValidationIssues(
-	params: AskParams
+	params: AskParams,
+	options: ValidationOptions = {}
 ): AskValidationIssue[] {
 	const collector = createIssueCollector();
-	validateQuestions(params.questions, collector);
+	validateQuestions(params.questions, collector, options);
 	return collector.issues;
 }
 
@@ -47,12 +55,14 @@ function normalizeOption(option: AskOption): AskOption {
 		label: option.label.trim(),
 		description: option.description?.trim(),
 		preview: option.preview?.trim(),
+		...(option.freeform ? { freeform: true } : {}),
 	};
 }
 
 function validateQuestions(
 	questions: AskParams["questions"],
-	collector: IssueCollector
+	collector: IssueCollector,
+	options: ValidationOptions
 ) {
 	if (questions.length === 0) {
 		collector.add("questions", "At least one question is required");
@@ -61,7 +71,7 @@ function validateQuestions(
 
 	const questionIds = new Set<string>();
 	for (const [questionIndex, question] of questions.entries()) {
-		validateQuestion(question, questionIndex, questionIds, collector);
+		validateQuestion(question, questionIndex, questionIds, collector, options);
 	}
 }
 
@@ -69,7 +79,8 @@ function validateQuestion(
 	question: AskQuestionInput,
 	questionIndex: number,
 	questionIds: Set<string>,
-	collector: IssueCollector
+	collector: IssueCollector,
+	options: ValidationOptions
 ) {
 	const questionNumber = questionIndex + 1;
 	const questionPath = `questions[${questionIndex}]`;
@@ -114,6 +125,14 @@ function validateQuestion(
 		`Question ${questionNumber}: at least one option is required`
 	);
 
+	validateFreeformOptions(
+		question.options,
+		questionNumber,
+		collector,
+		`${questionPath}.options`,
+		options
+	);
+
 	const optionValues = new Set<string>();
 	for (const [optionIndex, option] of question.options.entries()) {
 		validateOption(
@@ -128,6 +147,32 @@ function validateQuestion(
 	}
 }
 
+function validateFreeformOptions(
+	options: AskOption[],
+	questionNumber: number,
+	collector: IssueCollector,
+	path: string,
+	validationOptions: ValidationOptions
+) {
+	const freeformCount = options.filter((option) => option.freeform).length;
+	if (freeformCount === 0) {
+		return;
+	}
+	if (!validationOptions.allowFreeform) {
+		collector.add(
+			path,
+			`Question ${questionNumber}: freeform options are only supported for /answer forms`
+		);
+		return;
+	}
+	if (freeformCount > 1 || options.length > 1) {
+		collector.add(
+			path,
+			`Question ${questionNumber}: freeform options must be the only option`
+		);
+	}
+}
+
 function validateOption(
 	option: AskOption,
 	optionIndex: number,

src/state/selectors.ts

@@ -13,6 +13,8 @@ const CUSTOM_OPTION: AskDisplayOption = {
 	isCustomOption: true,
 };
 
+const DEFAULT_FREEFORM_LABEL = "Type your answer:";
+
 export function getCurrentQuestion(state: AskState): AskQuestion | undefined {
 	return state.questions[state.activeTabIndex];
 }
@@ -34,9 +36,30 @@ export function getRenderableOptions(
 	if (!question) {
 		return [];
 	}
+	const freeformOption = getFreeformOption(question);
+	if (freeformOption) {
+		return [
+			{
+				...freeformOption,
+				label: DEFAULT_FREEFORM_LABEL,
+				isCustomOption: true,
+				isFreeformOnlyOption: true,
+			},
+		];
+	}
 	return [...question.options, CUSTOM_OPTION];
 }
 
+function getFreeformOption(
+	question: AskQuestion
+): AskDisplayOption | undefined {
+	if (question.options.length !== 1) {
+		return;
+	}
+	const option = question.options[0];
+	return option.freeform ? option : undefined;
+}
+
 export function getCurrentOption(
 	state: AskState
 ): AskDisplayOption | undefined {

src/types.ts

@@ -2,6 +2,7 @@ export type AskQuestionType = "single" | "multi" | "preview";
 
 export interface AskOption {
 	description?: string;
+	freeform?: boolean;
 	label: string;
 	preview?: string;
 	value: string;
@@ -147,6 +148,7 @@ export interface AskState {
 
 export interface AskDisplayOption extends AskOption {
 	isCustomOption?: boolean;
+	isFreeformOnlyOption?: boolean;
 }
 
 export type AskAction =

src/ui/controller.ts

@@ -52,8 +52,16 @@ type Tui = CustomCallbackArgs[0];
 type Theme = CustomCallbackArgs[1];
 type Keybindings = CustomCallbackArgs[2];
 type Done = (result: AskResult) => void;
+interface AskFlowOptions {
+	allowFreeform?: boolean;
+}
+
 type AskFlowParams = AskParams &
-	Pick<ExtensionContext, "cwd"> & { config: AskConfig; ctx: ExtensionContext };
+	Pick<ExtensionContext, "cwd"> & {
+		config: AskConfig;
+		ctx: ExtensionContext;
+		flowOptions: AskFlowOptions;
+	};
 
 interface AskFlowController {
 	config: AskConfig;
@@ -72,12 +80,19 @@ interface AskFlowController {
 
 export async function runAskFlow(
 	ctx: ExtensionContext,
-	params: AskParams
+	params: AskParams,
+	options: AskFlowOptions = {}
 ): Promise<AskResult> {
 	const store = getAskConfigStore();
 	const config = await store.getConfig();
 	return ctx.ui.custom<AskResult>((...args) =>
-		createAskFlowController(args, { ...params, config, cwd: ctx.cwd, ctx })
+		createAskFlowController(args, {
+			...params,
+			config,
+			cwd: ctx.cwd,
+			ctx,
+			flowOptions: options,
+		})
 	);
 }
 
@@ -97,7 +112,7 @@ function createAskFlowController(
 		done,
 		editor: createEditor(tui, theme, params.cwd),
 		settingsOpen: false,
-		state: createInitialState(params),
+		state: createInitialState(params, params.flowOptions),
 		suppressAutoInputForSelection: false,
 		pendingReviewShortcutActionIndex: undefined,
 		theme,
@@ -321,8 +336,10 @@ function submitEditor(controller: AskFlowController, value: string) {
 	if (nextState.activeTabIndex !== controller.state.activeTabIndex) {
 		clearDismissNotice(controller);
 	}
-	controller.state = maybeAutoSubmitState(nextState, controller.config);
+	controller.state = nextState;
-	controller.editor.setText("");
+	syncSelection(controller);
+	controller.state = maybeAutoSubmitState(controller.state, controller.config);
+	hydrateEditor(controller);
 	refresh(controller);
 	maybeFinish(controller);
 }

src/ui/render-question.ts

@@ -76,7 +76,7 @@ function renderStandardOption(
 
 	pushWrappedText(
 		lines,
-		`${row.index + 1}. ${row.prefix}${row.label}`,
+		formatOptionLabel(row),
 		context.width,
 		context.theme,
 		row.color,
@@ -179,11 +179,16 @@ function renderOptionDetail(
 	lines: string[],
 	detail: OptionDetailModel | undefined,
 	context: QuestionRenderContext,
-	options: { suppressLeadingGap?: boolean } = {}
+	options: { indent?: string; suppressLeadingGap?: boolean } = {}
 ) {
 	if (!detail) {
 		return;
 	}
+	const indent = options.indent ?? "     ";
+	const padding =
+		indent === " "
+			? UI_DIMENSIONS.editorContentPadding
+			: UI_DIMENSIONS.editorIndentedPadding;
 	if (detail.withGap && !options.suppressLeadingGap) {
 		lines.push("");
 	}
@@ -193,8 +198,8 @@ function renderOptionDetail(
 			editor: context.editor,
 			width: context.width,
 			theme: context.theme,
-			indent: "     ",
+			indent,
-			padding: UI_DIMENSIONS.editorIndentedPadding,
+			padding,
 			placeholder: detail.placeholder,
 		});
 		return;
@@ -205,7 +210,7 @@ function renderOptionDetail(
 			note: detail.text,
 			width: context.width,
 			theme: context.theme,
-			indent: "     ",
+			indent,
 		});
 		return;
 	}
@@ -215,8 +220,8 @@ function renderOptionDetail(
 		context.width,
 		context.theme,
 		"muted",
-		"     ",
+		indent,
-		"     "
+		indent
 	);
 }
 
@@ -244,29 +249,29 @@ function renderEditorWithIndent(args: {
 	});
 }
 
+function formatOptionLabel(row: OptionRowModel): string {
+	return row.isFreeformOnly
+		? row.label
+		: `${row.index + 1}. ${row.prefix}${row.label}`;
+}
+
 function renderInteractiveCustomOption(
 	lines: string[],
 	row: OptionRowModel,
 	context: QuestionRenderContext
 ) {
+	const indent = row.isFreeformOnly ? " " : row.pointer;
 	pushWrappedText(
 		lines,
-		`${row.index + 1}. ${row.prefix}${row.label}`,
+		formatOptionLabel(row),
 		context.width,
 		context.theme,
 		row.color,
-		row.pointer,
+		indent,
-		" ".repeat(visibleWidth(row.pointer))
+		" ".repeat(visibleWidth(indent))
 	);
-	renderEditorWithIndent({
+	renderOptionDetail(lines, row.detail, context, {
-		lines,
+		indent: row.isFreeformOnly ? " " : undefined,
-		editor: context.editor,
-		width: context.width,
-		theme: context.theme,
-		indent: "     ",
-		padding: UI_DIMENSIONS.editorIndentedPadding,
-		placeholder:
-			row.detail?.kind === "editor" ? row.detail.placeholder : "Type answer...",
 	});
 }
 

src/ui/view-models/question.ts

@@ -26,6 +26,7 @@ export interface OptionRowModel {
 	detail?: OptionDetailModel;
 	index: number;
 	isCustom: boolean;
+	isFreeformOnly: boolean;
 	label: string;
 	pointer: string;
 	prefix: string;
@@ -89,6 +90,7 @@ function buildOptionRowModel(
 	const answered = option.isCustomOption
 		? !!answer?.customText
 		: isOptionSelected(answer, option.value);
+	const pointer = getOptionPointer(option, selected);
 	return {
 		color: getOptionColor(answered, selected),
 		description: option.description,
@@ -98,13 +100,21 @@ function buildOptionRowModel(
 				: buildOptionDetailModel(context, option, selected),
 		index,
 		isCustom: !!option.isCustomOption,
+		isFreeformOnly: !!option.isFreeformOnlyOption,
 		label: option.label,
-		pointer: selected ? "❯ " : "  ",
+		pointer,
 		prefix: getOptionPrefix(question.type, option, answered),
 		selected,
 	};
 }
 
+function getOptionPointer(option: AskDisplayOption, selected: boolean): string {
+	if (option.isFreeformOnlyOption) {
+		return "";
+	}
+	return selected ? "❯ " : "  ";
+}
+
 function buildOptionDetailModel(
 	context: QuestionRenderContext,
 	option: AskDisplayOption | undefined,
@@ -139,11 +149,15 @@ function buildOptionDetailModel(
 		return {
 			kind: "editor",
 			placeholder: UI_TEXT.editorPlaceholderInput,
-			withGap: false,
+			withGap: !!option.isFreeformOnlyOption,
 		};
 	}
 	if (customText) {
-		return { kind: "custom-text", text: customText, withGap: false };
+		return {
+			kind: "custom-text",
+			text: customText,
+			withGap: !!option.isFreeformOnlyOption,
+		};
 	}
 	if (note) {
 		return { kind: "saved-note", text: note, withGap: false };

tests/answer-extraction.test.ts

@@ -0,0 +1,126 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+import {
+	collectExtractionBusinessIssues,
+	repairExtractionParams,
+	selectExtractionModel,
+} from "../src/answer-extraction.ts";
+
+function model(provider: string, id: string) {
+	return { provider, id } as never;
+}
+
+test("selectExtractionModel uses first configured model with auth", async () => {
+	const first = model("missing", "a");
+	const second = model("ok", "b");
+	const result = await selectExtractionModel(
+		{
+			model: model("fallback", "c"),
+			modelRegistry: {
+				find(provider: string, _id: string) {
+					if (provider === "missing") {
+						return first;
+					}
+					if (provider === "ok") {
+						return second;
+					}
+				},
+				getApiKeyAndHeaders(candidate: { provider: string }) {
+					if (candidate.provider === "ok") {
+						return Promise.resolve({ ok: true, apiKey: "key" });
+					}
+					return Promise.resolve({ ok: false, error: "no" });
+				},
+			},
+		} as never,
+		[
+			{ provider: "missing", id: "a" },
+			{ provider: "ok", id: "b" },
+		]
+	);
+
+	assert.deepEqual(result, {
+		model: second,
+		auth: { ok: true, apiKey: "key" },
+		usedFallback: false,
+	});
+});
+
+test("extraction business rules flag option spam", () => {
+	const issues = collectExtractionBusinessIssues({
+		questions: [
+			{
+				id: "q",
+				prompt: "Pick one",
+				options: [
+					{ value: "a", label: "A" },
+					{ value: "b", label: "B" },
+					{ value: "c", label: "C" },
+					{ value: "d", label: "D" },
+					{ value: "e", label: "E" },
+				],
+			},
+		],
+	});
+
+	assert.deepEqual(issues, ["questions[0].options has 5 items; max is 4"]);
+});
+
+test("repairExtractionParams caps options while preserving other", () => {
+	const repaired = repairExtractionParams({
+		questions: [
+			{
+				id: "q",
+				prompt: "Pick one",
+				options: [
+					{ value: "a", label: "A" },
+					{ value: "b", label: "B" },
+					{ value: "c", label: "C" },
+					{ value: "d", label: "D" },
+					{ value: "other", label: "Other" },
+				],
+			},
+		],
+	});
+
+	assert.deepEqual(
+		repaired.questions[0]?.options.map((option) => option.value),
+		["a", "b", "c", "other"]
+	);
+});
+
+test("repairExtractionParams drops generic conversational prompts", () => {
+	const repaired = repairExtractionParams({
+		questions: [
+			{
+				id: "generic",
+				prompt: "How can I help you today?",
+				options: [{ value: "help", label: "Help me" }],
+			},
+		],
+	});
+
+	assert.deepEqual(repaired.questions, []);
+});
+
+test("selectExtractionModel validates fallback model auth", async () => {
+	const fallback = model("fallback", "c");
+	const result = await selectExtractionModel(
+		{
+			model: fallback,
+			modelRegistry: {
+				find() {
+					return;
+				},
+				getApiKeyAndHeaders() {
+					return Promise.resolve({ ok: false, error: "no auth" });
+				},
+			},
+		} as never,
+		[]
+	);
+
+	assert.deepEqual(result, {
+		error: "No auth for fallback chat model: fallback/c.",
+	});
+});

tests/ask-payload-store.test.ts

@@ -0,0 +1,95 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+import { findLatestPayloadInCurrentBranch } from "../src/ask-payload-store.ts";
+import type { AskParams } from "../src/types.ts";
+
+const params: AskParams = {
+	questions: [
+		{
+			id: "goal",
+			prompt: "Goal?",
+			options: [{ value: "a", label: "A" }],
+		},
+	],
+};
+
+function custom(data: unknown) {
+	return { type: "custom", customType: "ask:payload", data };
+}
+
+function ctx(branch: unknown[]) {
+	return { sessionManager: { getBranch: () => branch } } as never;
+}
+
+test("branch payload lookup returns the latest valid matching source", () => {
+	const older = {
+		version: 1,
+		source: "tool",
+		params,
+		timestamp: 1,
+	};
+	const newer = {
+		version: 1,
+		source: "tool",
+		params: { ...params, title: "Newer" },
+		timestamp: 2,
+	};
+
+	const result = findLatestPayloadInCurrentBranch(
+		ctx([custom(older), custom(newer)]),
+		"tool"
+	);
+
+	assert.equal(result.data, newer);
+	assert.equal(result.invalidMatchFound, false);
+});
+
+test("branch payload lookup allows freeform only for answer extraction payloads", () => {
+	const freeformParams: AskParams = {
+		questions: [
+			{
+				id: "language",
+				prompt: "Which language?",
+				options: [{ value: "freeform", label: "Type answer", freeform: true }],
+			},
+		],
+	};
+
+	const answerPayload = {
+		version: 1,
+		source: "answer-extraction",
+		params: freeformParams,
+		timestamp: 1,
+	};
+	const toolPayload = {
+		version: 1,
+		source: "tool",
+		params: freeformParams,
+		timestamp: 2,
+	};
+
+	assert.equal(
+		findLatestPayloadInCurrentBranch(
+			ctx([custom(answerPayload)]),
+			"answer-extraction"
+		).data,
+		answerPayload
+	);
+	assert.equal(
+		findLatestPayloadInCurrentBranch(ctx([custom(toolPayload)]), "tool").data,
+		undefined
+	);
+});
+
+test("branch payload lookup ignores invalid stored payloads", () => {
+	const result = findLatestPayloadInCurrentBranch(
+		ctx([
+			custom({ version: 1, source: "tool", params: { questions: [] } }),
+			custom({ version: 1, source: "answer-extraction", params }),
+		]),
+		"tool"
+	);
+
+	assert.equal(result.data, undefined);
+	assert.equal(result.invalidMatchFound, true);
+});

tests/ask-tool.test.ts

@@ -24,19 +24,26 @@ const noop = () => {
 
 function registerMockTool() {
 	const tools: Record<string, unknown>[] = [];
+	const entries: Array<{ customType: string; data: unknown }> = [];
 	registerAskTool({
+		appendEntry(customType: string, data: unknown) {
+			entries.push({ customType, data });
+		},
 		registerTool(tool: unknown) {
 			tools.push(tool as Record<string, unknown>);
 		},
 	} as never);
-	return tools[0] as {
+	return {
-		execute: (...args: any[]) => Promise<any>;
+		entries,
-		renderCall: (args: unknown, theme: any) => { text: string };
+		tool: tools[0] as {
-		renderResult: (
+			execute: (...args: any[]) => Promise<any>;
-			result: any,
+			renderCall: (args: unknown, theme: any) => { text: string };
-			options: unknown,
+			renderResult: (
-			theme: any
+				result: any,
-		) => { text: string };
+				options: unknown,
+				theme: any
+			) => { text: string };
+		},
 	};
 }
 
@@ -61,8 +68,25 @@ function sampleParams(): AskParams {
 	};
 }
 
+test("ask tool stores valid payloads as soon as they are called", async () => {
+	const { entries, tool } = registerMockTool();
+	const params = sampleParams();
+
+	await tool.execute("call-1", params, undefined, noop, makeCtx(false));
+
+	assert.equal(entries.length, 1);
+	assert.equal(entries[0].customType, "ask:payload");
+	assert.deepEqual(entries[0].data, {
+		version: 1,
+		source: "tool",
+		params,
+		sourceEntryId: "call-1",
+		timestamp: (entries[0].data as { timestamp: number }).timestamp,
+	});
+});
+
 test("ask tool returns pending questions in non-interactive mode", async () => {
-	const tool = registerMockTool();
+	const { tool } = registerMockTool();
 
 	const result = await tool.execute(
 		"call-1",
@@ -90,7 +114,7 @@ test("ask tool returns pending questions in non-interactive mode", async () => {
 });
 
 test("ask tool includes custom answer fallback for preview questions", async () => {
-	const tool = registerMockTool();
+	const { tool } = registerMockTool();
 
 	const result = await tool.execute(
 		"call-1",
@@ -115,7 +139,7 @@ test("ask tool includes custom answer fallback for preview questions", async ()
 });
 
 test("ask tool rejects invalid payloads before UI opens with structured issues", async () => {
-	const tool = registerMockTool();
+	const { tool } = registerMockTool();
 
 	const result = await tool.execute(
 		"call-1",
@@ -155,7 +179,7 @@ test("ask tool rejects invalid payloads before UI opens with structured issues",
 });
 
 test("ask tool reports missing option values with structured issues", async () => {
-	const tool = registerMockTool();
+	const { tool } = registerMockTool();
 
 	const result = await tool.execute(
 		"call-1",
@@ -188,7 +212,7 @@ test("ask tool reports missing option values with structured issues", async () =
 });
 
 test("ask tool reports empty questions with structured issues", async () => {
-	const tool = registerMockTool();
+	const { tool } = registerMockTool();
 
 	const result = await tool.execute(
 		"call-1",
@@ -215,7 +239,7 @@ test("ask tool reports empty questions with structured issues", async () => {
 });
 
 test("ask tool reports invalid question types with structured issues", async () => {
-	const tool = registerMockTool();
+	const { tool } = registerMockTool();
 
 	const result = await tool.execute(
 		"call-1",
@@ -250,7 +274,7 @@ test("ask tool reports invalid question types with structured issues", async ()
 });
 
 test("ask tool reports preview validation with structured issues", async () => {
-	const tool = registerMockTool();
+	const { tool } = registerMockTool();
 
 	const result = await tool.execute(
 		"call-1",
@@ -285,7 +309,7 @@ test("ask tool reports preview validation with structured issues", async () => {
 });
 
 test("ask tool transcript renderers summarize call and cancelled result", () => {
-	const tool = registerMockTool();
+	const { tool } = registerMockTool();
 	const theme = {
 		bold: (text: string) => text,
 		fg: (_token: string, text: string) => text,

tests/config-store.test.ts

@@ -24,7 +24,8 @@ test("config store writes defaults when file is missing", async () => {
 
 	assert.deepEqual(result.config, DEFAULT_ASK_CONFIG);
 	assert.deepEqual(JSON.parse(await readFile(path, "utf-8")), {
-		schemaVersion: 1,
+		schemaVersion: 2,
+		answer: DEFAULT_ASK_CONFIG.answer,
 		behaviour: DEFAULT_ASK_CONFIG.behaviour,
 		keymaps: DEFAULT_ASK_CONFIG.keymaps,
 	});
@@ -47,7 +48,8 @@ test("config store writes full normalized config on save", async () => {
 
 	const content = await readFile(path, "utf-8");
 	assert.deepEqual(JSON.parse(content), {
-		schemaVersion: 1,
+		schemaVersion: 2,
+		answer: DEFAULT_ASK_CONFIG.answer,
 		behaviour: {
 			autoSubmitWhenAnsweredWithoutNotes: true,
 			confirmDismissWhenDirty: true,
@@ -143,7 +145,8 @@ test("config store migrates the legacy root config path into extensions", async
 	assert.equal(result.config.behaviour.showFooterHints, false);
 	await assert.rejects(readFile(legacyPath, "utf-8"));
 	assert.deepEqual(JSON.parse(await readFile(path, "utf-8")), {
-		schemaVersion: 1,
+		schemaVersion: 2,
+		answer: DEFAULT_ASK_CONFIG.answer,
 		behaviour: {
 			autoSubmitWhenAnsweredWithoutNotes: true,
 			confirmDismissWhenDirty: true,

tests/editor-state.test.ts

@@ -66,6 +66,91 @@ test("submitEditorDraft advances submitted custom answers", () => {
 	assert.equal(next.view.kind, "submit");
 });
 
+test("submitted answer can sync directly into next freeform input", () => {
+	let state = syncStateToSelection(
+		createInitialState(
+			{
+				questions: [
+					{
+						id: "q1",
+						prompt: "First?",
+						options: [
+							{ value: "freeform", label: "Type answer", freeform: true },
+						],
+					},
+					{
+						id: "q2",
+						prompt: "Second?",
+						options: [
+							{ value: "freeform", label: "Type answer", freeform: true },
+						],
+					},
+				],
+			},
+			{ allowFreeform: true }
+		)
+	);
+
+	state = submitEditorDraft(state, "first answer");
+	state = syncStateToSelection(state);
+
+	assert.equal(state.activeTabIndex, 1);
+	assert.deepEqual(state.view, { kind: "input", questionId: "q2" });
+});
+
+test("syncStateToSelection does not reopen answered freeform input", () => {
+	let state = syncStateToSelection(
+		createInitialState(
+			{
+				questions: [
+					{
+						id: "q1",
+						prompt: "First?",
+						options: [
+							{ value: "freeform", label: "Type answer", freeform: true },
+						],
+					},
+				],
+			},
+			{ allowFreeform: true }
+		)
+	);
+	state = submitEditorDraft(state, "already answered");
+	state = syncStateToSelection({
+		...state,
+		activeTabIndex: 0,
+		view: { kind: "navigate" },
+	});
+
+	assert.equal(state.view.kind, "navigate");
+	assert.equal(state.answers.q1.customText, "already answered");
+});
+
+test("tab navigation can sync from review back into freeform input", () => {
+	let state = createInitialState(
+		{
+			questions: [
+				{
+					id: "q1",
+					prompt: "First?",
+					options: [
+						{ value: "freeform", label: "Type answer", freeform: true },
+					],
+				},
+			],
+		},
+		{ allowFreeform: true }
+	);
+	state = { ...state, activeTabIndex: 1, view: { kind: "submit" } };
+	state = syncStateToSelection({
+		...state,
+		activeTabIndex: 0,
+		view: { kind: "navigate" },
+	});
+
+	assert.deepEqual(state.view, { kind: "input", questionId: "q1" });
+});
+
 test("submitEditorDraft saves notes without completing the ask flow", () => {
 	const state = enterQuestionNoteMode(
 		createInitialState({

tests/render-question.test.ts

@@ -189,6 +189,47 @@ test("selected multiline custom option renders full editor block", () => {
 	assert(!lines.some((line) => line.includes("first line second line")));
 });
 
+test("freeform-only question renders label without numbering or pointer and separates input", () => {
+	const state = createInitialState(
+		{
+			questions: [
+				{
+					id: "q1",
+					prompt: "Type one",
+					options: [
+						{ value: "freeform", label: "Type answer", freeform: true },
+					],
+				},
+			],
+		},
+		{ allowFreeform: true }
+	);
+
+	const lines: string[] = [];
+	renderQuestionScreen({
+		editor: mockEditor("", ["┌────┐", "", "└────┘"]),
+		lines,
+		options: getRenderableOptions(state.questions[0]),
+		question: state.questions[0],
+		state: { ...state, view: { kind: "input", questionId: "q1" } },
+		theme: mockTheme(),
+		width: 80,
+	});
+
+	const labelIndex = lines.findIndex((line) =>
+		line.includes("Type your answer:")
+	);
+	const inputIndex = lines.findIndex((line) => line.includes("Type answer..."));
+
+	assert.notEqual(labelIndex, -1);
+	assert.equal(inputIndex, labelIndex + 2);
+	assert(lines[labelIndex]?.startsWith(" "));
+	assert(lines[inputIndex]?.startsWith(" "));
+	assert(!lines[inputIndex]?.startsWith("     "));
+	assert(!lines.some((line) => line.includes("1. Type your answer:")));
+	assert(!lines.some((line) => line.includes("❯")));
+});
+
 test("preview questions also show the custom answer option", () => {
 	const state = createInitialState({
 		questions: [

tests/settings-list.test.ts

@@ -1,10 +1,14 @@
 import assert from "node:assert/strict";
 import test from "node:test";
 import { visibleWidth } from "@mariozechner/pi-tui";
+import { DEFAULT_ASK_CONFIG } from "../src/config/defaults.ts";
 import type { AskConfig } from "../src/config/schema.ts";
 import { AskSettingsList } from "../src/ui/settings-list.ts";
 
 const savedConfig: AskConfig = {
+	answer: {
+		...DEFAULT_ASK_CONFIG.answer,
+	},
 	behaviour: {
 		autoSubmitWhenAnsweredWithoutNotes: false,
 		confirmDismissWhenDirty: true,

tests/state.test.ts

@@ -6,6 +6,9 @@ const DUPLICATE_QUESTION_ID_RE = /duplicate question id "scope"/;
 const DUPLICATE_OPTION_VALUE_RE = /duplicate option value "small"/;
 const PREVIEW_REQUIRED_RE =
 	/preview questions require preview text for every option/;
+const FREEFORM_UNSUPPORTED_RE =
+	/freeform options are only supported for \/answer forms/;
+const FREEFORM_MIXED_RE = /freeform options must be the only option/;
 
 import { toAskResult } from "../src/state/result.ts";
 import { getRenderableOptions } from "../src/state/selectors.ts";
@@ -76,6 +79,68 @@ test("normalize defaults via initial state", () => {
 	);
 });
 
+test("answer freeform questions render only the custom input option", () => {
+	const state = createInitialState(
+		{
+			questions: [
+				{
+					id: "language",
+					prompt: "Which language do you want to learn?",
+					options: [
+						{ value: "freeform", label: "Type answer", freeform: true },
+					],
+				},
+			],
+		},
+		{ allowFreeform: true }
+	);
+
+	const options = getRenderableOptions(state.questions[0]);
+	assert.equal(options.length, 1);
+	assert.equal(options[0]?.label, "Type your answer:");
+	assert.equal(options[0]?.isCustomOption, true);
+});
+
+test("freeform questions are rejected unless explicitly allowed", () => {
+	assert.throws(
+		() =>
+			createInitialState({
+				questions: [
+					{
+						id: "language",
+						prompt: "Which language do you want to learn?",
+						options: [
+							{ value: "freeform", label: "Type answer", freeform: true },
+						],
+					},
+				],
+			}),
+		FREEFORM_UNSUPPORTED_RE
+	);
+});
+
+test("freeform questions cannot mix normal options", () => {
+	assert.throws(
+		() =>
+			createInitialState(
+				{
+					questions: [
+						{
+							id: "language",
+							prompt: "Which language do you want to learn?",
+							options: [
+								{ value: "freeform", label: "Type answer", freeform: true },
+								{ value: "french", label: "French" },
+							],
+						},
+					],
+				},
+				{ allowFreeform: true }
+			),
+		FREEFORM_MIXED_RE
+	);
+});
+
 test("preview questions keep their type and option previews", () => {
 	const state = createInitialState({
 		questions: [