project: bump version to 0.1.1

set input-meta/output-meta on and convert-meta off (best-effort; full
grapheme delete still depends on terminal/readline — see CPython #142162).

CLAUDE.md

@@ -75,9 +75,9 @@ Module-level `@tool` handlers. Call `set_workspace_root()` before use.
 - **`process`**: `run_command` (argv, no shell, timeout, optional stdin/env); PTY `open_pty` / `read_pty` / `write_pty` / `close_pty` (POSIX: `pty`+`fork`; Windows: ConPTY via `pywinpty` env marker dep).
 - PTY: backend in `pty_backend.py`; structured status (`alive`/`exit_code`/`data`); `read_pty(..., until=)`; session limit/idle TTL/output budget; close terminate→kill.
 - CLI limit hooks: interactive confirm to raise tool-loop rounds, PTY session cap, or PTY output budget.
-- Destructive confirms: `classify_danger` + `ToolRegistry(on_confirm=…)`; CLI default deny; config `confirm_destructive` / `path_denylist`.
+- Destructive confirms: `classify_danger` + `ToolRegistry(on_confirm=…)`; CLI default deny; config `confirm_destructive` / `path_denylist`. Session YOLO: `/yolo on|off|once` and `--yes` (skip soft confirms; hard denylists unchanged; `once` expires after the next user turn).
 - **`vcs`**: read-only VCS tools (`vcs_kind` / `vcs_status` / `vcs_diff` / `vcs_log` / `vcs_branch`) via `VcsBackend` protocol; **git** implemented; detectors are pluggable for other systems.
-- **`chat`**: human prompts as tools — `ask_user`, `choose_user`, `form_user` (shared `prompting` core).
+- **`chat`**: human prompts as tools — `ask_user_line` / `ask_user_choice` / `ask_user_form` (shared `prompting` core).
 - **`DEFAULT_TOOLS`**: file + process + vcs + chat tool list for a `ToolRegistry`.
 
 ### Prompting (`prompting.py`)
@@ -88,8 +88,8 @@ Shared interactive I/O: `ask` / `choose` / `form` / `confirm` with pluggable bac
 
 Click app + readline REPL. Entry: `plyngent` / `python -m plyngent`.
 
-- **`plyngent chat`**: provider/model (flags or interactive; Tab via readline in `prompting`); sessions store `provider_name`/`model` and restore on resume; SQLite via `[database]` (file DB under user data if unset/`:memory:`); workspace-bound; resume latest for cwd/`--workspace` by default (`--new` / `--session`). One-shot: `-p/--prompt` and non-TTY stdin; exit codes 0/1/2/3; `--yes`, `--stream/--no-stream`, `--quiet`. Root `--log-level`.
-- Slash: Click group in `cli/slash.py` + `awaitlet` for async work; Tab completer from registry + ParamType `shell_complete`. Multiline `"""` … `"""`; `/edit` via `$EDITOR`.
+- **`plyngent chat`**: provider/model (flags or interactive; Tab via readline in `prompting`); sessions store `provider_name`/`model` and restore on resume; SQLite via `[database]` (file DB under user data if unset/`:memory:`); workspace-bound; resume latest for cwd/`--workspace` by default (`--new` / `--session`). One-shot: `-p/--prompt` and non-TTY stdin; exit codes 0/1/2/3; `--yes` (YOLO on), `--stream/--no-stream`, `--quiet`. Root `--log-level`.
+- Slash: Click group in `cli/slash.py` + `awaitlet` for async work; Tab completer from registry + ParamType `shell_complete`. Multiline `"""` … `"""`; `/edit` via `$EDITOR`. `/yolo on|off|once` for soft destructive confirms.
 - Explicit `/resume` or `--session` from another workspace prompts: **keep** / **update** / **abort**.
 - Failed/cancelled turns: user kept; **committed tool rounds kept** (side effects not re-run on `/retry`); only unfinished assistant rolled back; Ctrl+C cancels; TTY confirms off-loop; auto-retry 10s/20s/30s then `/retry`.
 - **`plyngent providers`**, **`config path|edit`**. No providers + `$EDITOR` → optional edit then reload.
@@ -112,7 +112,7 @@ Basedpyright `recommended`. Ruff includes `ANN` (private return types `ANN202` i
 - **Phase G (CLI polish + hardening)** — single-user only; multi-tenant stays Phase H.
 
   **Done**
-  - G0: `prompting` (`ask`/`choose`/`form`/`confirm`) + chat tools (`ask_user`/`choose_user`/`form_user`)
+  - G0: `prompting` (`ask`/`choose`/`form`/`confirm`) + chat tools (`ask_user_line`/`ask_user_choice`/`ask_user_form`)
   - G1: `ReasoningDeltaEvent`, `/stream`, `/verbose`
   - G2: `/rename`, `/delete` (confirm), `/export md|json`
   - G2.5: Click slash registry (`cli/slash.py`) + `awaitlet` for sync Click / async memory; auto `/help`; completer from `slash.list_commands`

README.md

@@ -152,7 +152,7 @@ plyngent chat --session 3
 | `--max-rounds` | Tool-loop rounds per turn (default 32) |
 | `--stream` / `--no-stream` | Streaming deltas (default on) |
 | `--quiet` | Less status on stderr |
-| `--yes` | Allow destructive tools without confirm (also for one-shot) |
+| `--yes` | YOLO on: skip destructive-tool confirms for this process |
 | `--log-level` | On the root CLI: `DEBUG`, `INFO`, `WARNING`, … |
 
 Sessions resume the **most recently updated** session for the current workspace unless you pass `--new` or `--session`. Each session remembers the last **provider** and **model** (restored on resume so you are not re-prompted).
@@ -192,6 +192,7 @@ Type `/help` in the REPL for the live list. Common ones:
 | `/export [md\|json] [path]` | Transcript from DB (no secrets) |
 | `/compact` | Soft-compact + model summary into a **new** session |
 | `/stream` `/verbose` `/markdown` `/tools` `/rounds` | Toggles and limits |
+| `/yolo [on\|off\|once]` | Soft destructive confirms: sticky skip, off, or next turn only |
 
 | `/retry` | Re-run incomplete last user turn (after error/cancel) |
 | `/provider` `/model` | Switch without restarting |
@@ -209,13 +210,13 @@ User messages are saved immediately. On API error or Ctrl+C, partial assistant/t
 
 ## Tools (when enabled)
 
-Default registry: file ops, `run_command` / PTY (POSIX openpty; Windows ConPTY via pywinpty), read-only VCS (git), and human prompts (`ask_user` / `choose_user` / `form_user`).
+Default registry: file ops, `run_command` / PTY (POSIX openpty; Windows ConPTY via pywinpty), read-only VCS (git), and human prompts (`ask_user_line` / `ask_user_choice` / `ask_user_form`).
 
 Safety defaults:
 
 - Paths stay under the workspace; optional `path_denylist` substrings.
 - Command basename denylist (e.g. dangerous shells/utilities).
-- Destructive tools (delete/move/overwrite) can require confirm (`confirm_destructive`; default deny in non-TTY).
+- Destructive tools (delete/move/overwrite) can require confirm (`confirm_destructive`; default deny in non-TTY). Override for the session with `/yolo on|off|once` or startup `--yes` (path/command denylists still apply).
 - PTY sessions: caps, idle TTL, output budget; master FD is non-inheritable; sessions closed on chat exit.
 
 ## Usage / context (CLI)

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "plyngent"
-version = "0.1.0"
+version = "0.1.1"
 description = "LLM chat and agent toolkit"
 authors = [
   { name = "HivertMoZara", email = "worldmozara@163.com" },

src/plyngent/agent/tools.py

@@ -191,6 +191,11 @@ class ToolRegistry:
     def __len__(self) -> int:
         return len(self._tools)
 
+    @property
+    def soft_confirm(self) -> bool:
+        """True when dangerous tools are gated by ``on_confirm``."""
+        return self._danger is not None and self._on_confirm is not None
+
     async def _invoke(self, definition: ToolDefinition, args: dict[str, object]) -> str:
         try:
             result = definition.handler(**args)

src/plyngent/cli/app.py

@@ -171,13 +171,16 @@ async def _run_oneshot(state: ReplState, prompt_text: str) -> int:
     try:
         ok = await run_user_text_with_retries(state.agent, prompt_text, delays=())
     except asyncio.CancelledError:
+        state.expire_yolo_once(quiet=True)
         return EXIT_CANCELLED
     except KeyboardInterrupt:
+        state.expire_yolo_once(quiet=True)
         return EXIT_CANCELLED
+    state.expire_yolo_once(quiet=True)
     return EXIT_OK if ok else EXIT_TURN_FAILED
 
 
-async def _run_chat(  # noqa: C901, PLR0912 — chat orchestration
+async def _run_chat(  # noqa: C901, PLR0912, PLR0915 — chat orchestration
     *,
     config_path: Path | None,
     provider_name: str | None,
@@ -206,7 +209,10 @@ async def _run_chat(  # noqa: C901, PLR0912 — chat orchestration
             _warn_recoverable_providers(store.recoverable_providers)
 
     _setup_workspace_and_hooks(store, workspace, interactive=interactive)
-    confirm_destructive: bool | None = False if yes else None
+    # --yes forces sticky YOLO; else derive from config.confirm_destructive.
+    from plyngent.cli.state import YoloMode
+
+    yolo: YoloMode | None = "on" if yes else None
 
     memory = await MemoryStore.open(_database_config(store, quiet=quiet or oneshot))
     try:
@@ -239,8 +245,27 @@ async def _run_chat(  # noqa: C901, PLR0912 — chat orchestration
                 preferred_model=preferred_model,
                 interactive=interactive,
             )
-            model_id = select_model(provider, preferred=preferred_model, interactive=interactive)
-            _ = create_client(provider)
+            # Build client early so we can always try GET /models for remote-first lists.
+            from plyngent.cli.models_source import (
+                client_supports_models,
+                fetch_remote_model_ids,
+                model_choices_for_provider,
+            )
+
+            client = create_client(provider)
+            remote_ids: list[str] | None = None
+            try:
+                if client_supports_models(client):
+                    remote_ids = await fetch_remote_model_ids(client)
+            except (RuntimeError, TypeError, OSError, ValueError):
+                remote_ids = None
+            choices = model_choices_for_provider(provider, remote_ids=remote_ids)
+            model_id = select_model(
+                provider,
+                preferred=preferred_model,
+                interactive=interactive,
+                choices=choices,
+            )
         except ProviderNotSupportedError as exc:
             raise click.ClickException(str(exc)) from exc
 
@@ -255,8 +280,11 @@ async def _run_chat(  # noqa: C901, PLR0912 — chat orchestration
             max_rounds=max_rounds,
             stream_enabled=stream,
             interactive_limits=interactive,
-            confirm_destructive=confirm_destructive,
+            yolo=yolo,
         )
+        # Seed remote model cache from startup fetch so Tab/complete stays warm.
+        if remote_ids is not None:
+            state.seed_remote_models(remote_ids)
         if not quiet and not oneshot:
             click.secho(f"workspace: {state.workspace}", fg="bright_black", err=True)
 
@@ -367,7 +395,7 @@ def main(ctx: click.Context, log_level: str) -> None:
     "--yes",
     is_flag=True,
     default=False,
-    help="Allow destructive tools without confirm (one-shot / non-interactive).",
+    help="Enable YOLO: skip destructive-tool confirms (sticky for this process).",
 )
 @click.option(
     "--quiet",

src/plyngent/cli/display.py

@@ -4,7 +4,7 @@ import contextlib
 import os
 import sys
 from contextvars import ContextVar
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Literal
 
 import click
 
@@ -32,6 +32,8 @@ _TOOL_ARGS_PREVIEW = 80
 _verbose_tool_results: ContextVar[bool] = ContextVar("verbose_tool_results", default=False)
 _markdown_enabled: ContextVar[bool] = ContextVar("markdown_enabled", default=True)
 
+type StreamSource = Literal["reasoning", "assistant"]
+
 
 def set_verbose_tool_results(enabled: bool) -> None:  # noqa: FBT001
     """Set whether tool results print in full (True) or as a short preview."""
@@ -87,26 +89,40 @@ def _clear_streamed_lines(line_count: int) -> None:
 
 
 def _line_count_for_clear(label: str, body: str) -> int:
-    """Approximate terminal lines used by ``label + body`` for cursor erase."""
+    """Approximate terminal lines used by ``label\\n + body`` for cursor erase."""
     if not body and not label:
         return 0
-    text = label + body
-    # +1 for trailing newline after the stream block in render_events.
+    # Label is on its own line; body may contain newlines.
+    text = f"{label}\n{body}" if label else body
     return text.count("\n") + 1
 
 
-def print_markdown(text: str, *, label: str = "assistant: ") -> None:
-    """Render *text* as markdown via Rich, with a cyan label."""
+def print_markdown(text: str, *, label: str = "assistant:") -> None:
+    """Render *text* as markdown via Rich; *label* on its own line when set."""
     from rich.console import Console
     from rich.markdown import Markdown
     from rich.text import Text
 
     console = Console(file=sys.stdout, highlight=False)
     if label:
-        console.print(Text(label, style="cyan"), end="")
+        console.print(Text(label, style="cyan"))
     console.print(Markdown(text))
 
 
+def _flush_assistant_markdown(body: str, *, pretty: bool) -> None:
+    """Replace the plain assistant stream with markdown when enabled."""
+    if not body.strip():
+        click.echo()
+        return
+    if pretty:
+        lines = _line_count_for_clear("assistant:", body)
+        _clear_streamed_lines(lines)
+        print_markdown(body, label="assistant:")
+        click.echo()
+    else:
+        click.echo()
+
+
 async def render_events(  # noqa: C901, PLR0912, PLR0915
     events: AsyncIterator[AgentEvent],
     *,
@@ -115,38 +131,65 @@ async def render_events(  # noqa: C901, PLR0912, PLR0915
 ) -> None:
     """Print agent events to the terminal (text deltas stream as they arrive).
 
-    When markdown is enabled and stdout is a TTY, the plain streamed assistant
-    text is replaced at end-of-turn with a Rich markdown render.
+    Assistant and reasoning each start on a new line after their label. When the
+    content source changes (reasoning ↔ assistant, or tools/errors), the
+    assistant markdown buffer is flushed so streams do not mix and Rich can
+    re-render completed assistant segments.
     """
     show_full = get_verbose_tool_results() if verbose is None else verbose
     use_markdown = get_markdown_enabled() if markdown is None else markdown
     pretty = bool(use_markdown and markdown_render_available())
 
-    printed_reasoning = False
-    printed_text = False
+    source: StreamSource | None = None
     assistant_buf: list[str] = []
-    # Tools/errors after assistant text: skip replace (would erase tool lines).
-    interrupted_by_other = False
+    printed_reasoning = False
+    printed_assistant = False
+
+    def flush_assistant() -> None:
+        nonlocal source, assistant_buf, printed_assistant
+        if source != "assistant" and not assistant_buf:
+            return
+        body = "".join(assistant_buf)
+        assistant_buf = []
+        if printed_assistant:
+            _flush_assistant_markdown(body, pretty=pretty)
+        printed_assistant = False
+        if source == "assistant":
+            source = None
+
+    def begin_reasoning() -> None:
+        nonlocal source, printed_reasoning
+        if source == "reasoning":
+            return
+        if source == "assistant":
+            flush_assistant()
+        click.echo()
+        click.secho("reasoning:", fg="bright_black")
+        source = "reasoning"
+        printed_reasoning = True
+
+    def begin_assistant() -> None:
+        nonlocal source, printed_assistant
+        if source == "assistant":
+            return
+        if source == "reasoning":
+            click.echo()  # end reasoning stream line
+            source = None
+        click.echo()
+        click.secho("assistant:", fg="cyan")
+        source = "assistant"
+        printed_assistant = True
 
     async for event in events:
         if isinstance(event, ReasoningDeltaEvent):
-            if not printed_reasoning:
-                click.echo()
-                click.secho("reasoning: ", fg="bright_black", nl=False)
-                printed_reasoning = True
+            begin_reasoning()
             _echo_stream(event.content)
         elif isinstance(event, TextDeltaEvent):
-            if printed_reasoning and not printed_text:
-                click.echo()
-            if not printed_text:
-                click.echo()
-                click.secho("assistant: ", fg="cyan", nl=False)
-                printed_text = True
+            begin_assistant()
             assistant_buf.append(event.content)
             _echo_stream(event.content)
         elif isinstance(event, ToolCallEvent):
-            if printed_text:
-                interrupted_by_other = True
+            flush_assistant()
             call = event.tool_call
             if isinstance(call, AssistantFunctionToolCall):
                 args = _preview(call.function.arguments, _TOOL_ARGS_PREVIEW)
@@ -154,8 +197,7 @@ async def render_events(  # noqa: C901, PLR0912, PLR0915
             else:
                 click.secho(f"\n[tool] custom id={call.id}", fg="yellow")
         elif isinstance(event, ToolResultEvent):
-            if printed_text:
-                interrupted_by_other = True
+            flush_assistant()
             content = event.message.content
             if show_full:
                 click.secho(f"[tool ok]\n{content}", fg="magenta")
@@ -164,8 +206,7 @@ async def render_events(  # noqa: C901, PLR0912, PLR0915
                 one_line = preview.replace("\n", " ")
                 click.secho(f"[tool ok] {one_line}", fg="magenta")
         elif isinstance(event, ErrorEvent):
-            if printed_text:
-                interrupted_by_other = True
+            flush_assistant()
             suffix = ""
             if event.source:
                 suffix += f" source={event.source}"
@@ -173,15 +214,13 @@ async def render_events(  # noqa: C901, PLR0912, PLR0915
                 suffix += " (fatal)"
             click.secho(f"\n[error]{suffix} {event.message}", fg="bright_red")
         elif isinstance(event, CancelledEvent):
-            if printed_text:
-                interrupted_by_other = True
+            flush_assistant()
             if event.reason:
                 click.secho(f"\n[cancelled] {event.reason}", fg="yellow")
             else:
                 click.secho("\n[cancelled]", fg="yellow")
         elif isinstance(event, MaxRoundsEvent):
-            if printed_text:
-                interrupted_by_other = True
+            flush_assistant()
             if event.continued:
                 click.secho(
                     f"\n[max rounds {event.rounds} reached — continuing with a higher allowance]",
@@ -195,21 +234,9 @@ async def render_events(  # noqa: C901, PLR0912, PLR0915
             # AssistantMessageEvent — text already shown via TextDeltaEvent.
             _ = event
 
-    full_assistant = "".join(assistant_buf)
-    if pretty and printed_text and full_assistant.strip() and not interrupted_by_other:
-        # Replace plain stream with markdown (assistant block only).
-        lines = _line_count_for_clear("assistant: ", full_assistant)
-        # Also account for blank line before label when reasoning was shown.
-        if printed_reasoning:
-            lines += 1
-        _clear_streamed_lines(lines)
-        if printed_reasoning:
-            # Reasoning already printed above; leave it and only re-print assistant.
-            pass
-        print_markdown(full_assistant, label="assistant: ")
-        click.echo()
-        return
-
-    if printed_text or printed_reasoning:
+    # End-of-turn: flush any open assistant segment; close reasoning stream.
+    if assistant_buf or printed_assistant:
+        flush_assistant()
+    elif printed_reasoning:
         click.echo()
     click.echo()

src/plyngent/cli/models_source.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import inspect
-from typing import TYPE_CHECKING, Protocol, cast, runtime_checkable
+from typing import TYPE_CHECKING, Literal, Protocol, cast, runtime_checkable
 
 if TYPE_CHECKING:
     from collections.abc import Iterable, Sequence
@@ -11,6 +11,8 @@ if TYPE_CHECKING:
 # Cache remote catalog this long (seconds) unless /models --refresh.
 DEFAULT_MODELS_CACHE_TTL = 300.0
 
+type ModelListPrefer = Literal["remote", "union", "config"]
+
 
 @runtime_checkable
 class SupportsModels(Protocol):
@@ -25,12 +27,30 @@ def config_model_ids(provider: Provider) -> list[str]:
 def merge_model_choices(
     config_ids: Iterable[str],
     remote_ids: Iterable[str] | None = None,
+    *,
+    prefer: ModelListPrefer = "remote",
 ) -> list[str]:
-    """Union config and remote ids (sorted, unique)."""
-    merged: set[str] = {i for i in config_ids if i}
-    if remote_ids is not None:
-        merged.update(i for i in remote_ids if i)
-    return sorted(merged)
+    """Merge config and remote model ids.
+
+    *prefer*:
+    - ``remote`` (default): remote catalog first (sorted), then config-only ids
+    - ``union``: sorted unique union
+    - ``config``: config first, then remote-only ids
+    """
+    config_list = [i for i in config_ids if i]
+    remote_list = [i for i in (remote_ids or ()) if i]
+    if not remote_list:
+        return sorted(set(config_list))
+    if prefer == "union":
+        return sorted(set(config_list) | set(remote_list))
+    remote_sorted = sorted(set(remote_list))
+    config_only = sorted(set(config_list) - set(remote_sorted))
+    if prefer == "remote":
+        return [*remote_sorted, *config_only]
+    # config first
+    config_sorted = sorted(set(config_list))
+    remote_only = sorted(set(remote_list) - set(config_sorted))
+    return [*config_sorted, *remote_only]
 
 
 def client_supports_models(client: object) -> bool:
@@ -57,6 +77,7 @@ def model_choices_for_provider(
     provider: Provider,
     *,
     remote_ids: Sequence[str] | None = None,
+    prefer: ModelListPrefer = "remote",
 ) -> list[str]:
-    """Config plus remote catalog for selection / Tab complete."""
-    return merge_model_choices(config_model_ids(provider), remote_ids)
+    """Config plus remote catalog for selection / Tab complete (remote-first)."""
+    return merge_model_choices(config_model_ids(provider), remote_ids, prefer=prefer)

src/plyngent/cli/readline_setup.py

@@ -77,6 +77,27 @@ def bind_tab_complete(readline_mod: object) -> None:
         _ = parse("python:bind ^I rl_complete")
 
 
+def bind_utf8_input(readline_mod: object) -> None:
+    """Best-effort 8-bit / UTF-8 settings for GNU readline (CJK backspace).
+
+    GNU readline can mishandle wide characters when meta conversion is on.
+    These binds are no-ops or ignored on libedit. Full grapheme editing still
+    depends on the terminal and readline version (see CPython #142162).
+    """
+    parse = getattr(readline_mod, "parse_and_bind", None)
+    if not callable(parse):
+        return
+    for cmd in (
+        "set input-meta on",
+        "set output-meta on",
+        "set convert-meta off",
+        "set enable-meta-key on",
+        "set horizontal-scroll-mode off",
+    ):
+        with contextlib.suppress(Exception):
+            _ = parse(cmd)
+
+
 def setup_readline(state: ReplState) -> None:
     """Configure Tab completion and persistent history when readline is available."""
     try:
@@ -85,6 +106,7 @@ def setup_readline(state: ReplState) -> None:
         return
 
     bind_tab_complete(readline)
+    bind_utf8_input(readline)
     # Treat path-like chars as part of a token so /help completes as one word.
     readline.set_completer_delims(" \t\n")
     readline.set_completer(build_completer(state))

src/plyngent/cli/repl.py

@@ -25,12 +25,14 @@ def _echo_user(text: str) -> None:
 async def run_repl(state: ReplState) -> None:
     """Interactive chat loop with readline editing, history, and Tab completion."""
     setup_readline(state)
+    yolo = state.effective_yolo()
+    yolo_part = f"  yolo={yolo}" if yolo != "off" else ""
     click.echo(
         f"plyngent chat  provider={state.provider_name}  model={state.model}  "
         f"session={state.session_id}  tools={'on' if state.tools_enabled else 'off'}  "
         f"rounds={state.max_rounds}  messages={len(state.agent.messages)}  "
         f"stream={'on' if state.agent.stream else 'off'}  "
-        f"verbose={'on' if state.verbose else 'off'}"
+        f"verbose={'on' if state.verbose else 'off'}{yolo_part}"
     )
     click.echo('Type /help for commands. Multiline: """ … """. Empty line is ignored.')
 
@@ -52,8 +54,14 @@ async def run_repl(state: ReplState) -> None:
                 text = state.pending_user_text
                 state.pending_user_text = None
                 _echo_user(text)
+                try:
                     _ = await run_user_text_with_retries(state.agent, text)
+                finally:
+                    state.expire_yolo_once()
             continue
 
         _echo_user(entry)
+        try:
             _ = await run_user_text_with_retries(state.agent, entry)
+        finally:
+            state.expire_yolo_once()

src/plyngent/cli/retry.py

@@ -17,8 +17,23 @@ if TYPE_CHECKING:
     from plyngent.agent import AgentEvent
     from plyngent.agent.chat import ChatAgent
 
-# Wait before retry attempt 1, 2, and 3 (after the first failure).
-DEFAULT_RETRY_DELAYS_SECONDS: tuple[float, ...] = (10.0, 20.0, 30.0)
+# Auto-retry budget after the first failure (10 attempts by default).
+DEFAULT_MAX_AUTO_RETRIES = 10
+# First four waits; each further wait is previous + 10s.
+_RETRY_BASE_DELAYS_SECONDS: tuple[float, ...] = (5.0, 10.0, 15.0, 20.0)
+
+
+def default_retry_delays(max_retries: int = DEFAULT_MAX_AUTO_RETRIES) -> tuple[float, ...]:
+    """Build wait times: 5, 10, 15, 20, then +10s each step, length *max_retries*."""
+    if max_retries <= 0:
+        return ()
+    delays: list[float] = list(_RETRY_BASE_DELAYS_SECONDS)
+    while len(delays) < max_retries:
+        delays.append(delays[-1] + 10.0)
+    return tuple(delays[:max_retries])
+
+
+DEFAULT_RETRY_DELAYS_SECONDS: tuple[float, ...] = default_retry_delays()
 _PREVIEW_LEN = 80
 
 

src/plyngent/cli/slash.py

@@ -23,13 +23,14 @@ from plyngent.runtime import ProviderNotSupportedError
 if TYPE_CHECKING:
     from collections.abc import Awaitable, Sequence
 
-    from plyngent.cli.state import ReplState
+    from plyngent.cli.state import ReplState, YoloMode
     from plyngent.lmproto.openai_compatible.model import AnyChatMessage
 
 _DEFAULT_HISTORY_LINES = 20
 _CONTENT_PREVIEW = 200
 _COMPACT_PREVIEW = 400
 _ON_OFF_CHOICES = ("on", "off")
+_YOLO_MODE_CHOICES = ("on", "off", "once")
 _EXPORT_FORMAT_CHOICES = ("md", "json")
 
 HELP_FOOTER = (
@@ -38,7 +39,7 @@ HELP_FOOTER = (
     "works after resume, not only via readline history). Auto-retry: 10s/20s/30s.\n"
     "\n"
     "Tab completes slash commands and some arguments (provider, model, tools,\n"
-    "stream, verbose, export). Use --session ID or /resume to continue a prior\n"
+    "stream, verbose, yolo, export). Use --session ID or /resume to continue a prior\n"
     "chat after restart.\n"
     "\n"
     'Multiline: start a message with """ then end a later line with """.\n'
@@ -86,6 +87,30 @@ class OnOffParam(click.ParamType[bool]):
 ON_OFF = OnOffParam()
 
 
+class YoloModeParam(click.ParamType[str]):
+    """Accept on|off|once for soft destructive-tool confirms."""
+
+    name: str = "yolo_mode"
+
+    @override
+    def convert(self, value: Any, param: click.Parameter | None, ctx: click.Context | None) -> str:
+        if isinstance(value, str) and value in _YOLO_MODE_CHOICES:
+            return value
+        token = str(value).strip().lower()
+        if token in _YOLO_MODE_CHOICES:
+            return token
+        msg = "expected on, off, or once"
+        raise click.BadParameter(msg, ctx=ctx, param=param)
+
+    @override
+    def shell_complete(self, ctx: click.Context, param: click.Parameter, incomplete: str) -> list[CompletionItem]:
+        del ctx, param
+        return _filter_choices(incomplete, _YOLO_MODE_CHOICES)
+
+
+YOLO_MODE = YoloModeParam()
+
+
 class ExportFormatParam(click.ParamType[str]):
     """First token of /export: md|json (or a path if not a format)."""
 
@@ -355,7 +380,8 @@ def status_cmd(state: ReplState) -> None:
         f"tools={'on' if state.tools_enabled else 'off'}  "
         f"rounds={state.max_rounds}  "
         f"stream={'on' if state.agent.stream else 'off'}  "
-        f"verbose={'on' if state.verbose else 'off'}\n"
+        f"verbose={'on' if state.verbose else 'off'}  "
+        f"yolo={state.effective_yolo()}\n"
         f"context_tokens={ctx_tilde}{ctx_tokens}/{ctx_budget} ({ctx_tag})  "
         f"context_chars={ctx_chars}  "
         f"tool_result_max={state.agent.max_tool_result_chars}\n"
@@ -581,7 +607,8 @@ def provider_cmd(state: ReplState, name: str | None) -> None:
         state.provider_name = pname
         state.provider = provider
         state.rebuild_client()
-        choices = _await(state.merged_model_choices(refresh=False))
+        # Always request remote catalog for the new provider (bypass stale cache).
+        choices = _await(state.merged_model_choices(refresh=True))
         if prev_model and (prev_model in choices or prev_model in provider.models):
             state.model = prev_model
         else:
@@ -612,11 +639,12 @@ def provider_cmd(state: ReplState, name: str | None) -> None:
 @click.option("--refresh", is_flag=True, help="Bypass cache and re-fetch GET /models.")
 @click.pass_obj
 def models_cmd(state: ReplState, *, refresh: bool) -> None:
-    """List models (config plus remote GET /models)."""
+    """List models (remote-first, plus config-only ids). Always tries GET /models."""
+    del refresh  # always re-fetch; flag kept for CLI compatibility / docs
     remote: list[str] | None = None
     remote_err: str | None = None
     try:
-        remote = _await(state.ensure_remote_models(refresh=refresh))
+        remote = _await(state.ensure_remote_models(refresh=True))
     except (RuntimeError, TypeError, OSError, ValueError) as exc:
         remote_err = str(exc)
         remote = state.cached_remote_models()
@@ -643,10 +671,10 @@ def models_cmd(state: ReplState, *, refresh: bool) -> None:
         remote_set = set(remote or ())
         for mid in choices:
             tags: list[str] = []
-            if mid in config_ids:
-                tags.append("config")
             if mid in remote_set:
                 tags.append("remote")
+            if mid in config_ids:
+                tags.append("config")
             suffix = f"  ({', '.join(tags)})" if tags else ""
             mark = " *" if mid == state.model else ""
             click.echo(f"{mid}{mark}{suffix}")
@@ -655,7 +683,8 @@ def models_cmd(state: ReplState, *, refresh: bool) -> None:
         click.secho(f"remote list unavailable: {remote_err}", fg="yellow", err=True)
     elif remote is not None:
         click.echo(
-            f"({len(remote)} remote, {len(config_ids)} config; cache TTL {int(DEFAULT_MODELS_CACHE_TTL)}s)",
+            f"(remote-first: {len(remote)} remote, {len(config_ids)} config; "
+            f"cache TTL {int(DEFAULT_MODELS_CACHE_TTL)}s)",
             err=True,
         )
 
@@ -664,12 +693,12 @@ def models_cmd(state: ReplState, *, refresh: bool) -> None:
 @click.argument("model_id", required=False, type=MODEL_ID)
 @click.pass_obj
 def model_cmd(state: ReplState, model_id: str | None) -> None:
-    """Show or switch model (Tab: config plus cached remote)."""
+    """Show or switch model (Tab: remote-first plus config; live fetch on pick)."""
     if not model_id:
         click.echo(f"model={state.model}")
         return
     try:
-        choices = _await(state.merged_model_choices(refresh=False))
+        choices = _await(state.merged_model_choices(refresh=True))
         state.model = select_model(
             state.provider,
             preferred=model_id.strip(),
@@ -698,6 +727,24 @@ def tools_cmd(state: ReplState, enabled: bool | None) -> None:  # noqa: FBT001
     click.echo(f"tools={'on' if enabled else 'off'}")
 
 
+@slash.command("yolo")
+@click.argument("mode", required=False, type=YOLO_MODE, metavar="[on|off|once]")
+@click.pass_obj
+def yolo_cmd(state: ReplState, mode: str | None) -> None:
+    """Show or set YOLO mode for soft destructive-tool confirms.
+
+    ``off`` (default when config ``confirm_destructive`` is true): prompt on
+    delete/move/overwrite (deny in non-TTY). ``on``: skip confirms for the
+    process. ``once``: skip for the next user turn only, then return to ``off``.
+    Path/command denylists still apply. Omit the argument to print the value.
+    """
+    if mode is None:
+        click.echo(f"yolo={state.effective_yolo()}")
+        return
+    state.set_yolo(cast("YoloMode", mode))
+    click.echo(f"yolo={state.effective_yolo()}")
+
+
 @slash.command("stream")
 @click.argument("enabled", required=False, type=ON_OFF, metavar="[on|off]")
 @click.pass_obj

src/plyngent/cli/state.py

@@ -4,7 +4,7 @@ import contextlib
 import time
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import TYPE_CHECKING, cast
+from typing import TYPE_CHECKING, Literal, cast
 
 from plyngent.agent import ChatAgent, ChatClient, ToolRegistry
 from plyngent.agent.loop import DEFAULT_MAX_ROUNDS
@@ -25,6 +25,8 @@ if TYPE_CHECKING:
     from plyngent.memory import MemoryStore
     from plyngent.memory.database.schema import Session as SessionRow
 
+type YoloMode = Literal["off", "on", "once"]
+
 
 @dataclass
 class ReplState:
@@ -44,8 +46,9 @@ class ReplState:
     markdown_enabled: bool = True
     # One-shot / scripts: never prompt to raise tool-loop limits.
     interactive_limits: bool = True
-    # When False, skip destructive-tool confirms (e.g. --yes).
-    confirm_destructive: bool | None = None
+    # Soft destructive-tool confirms: None → derive from config.confirm_destructive.
+    # off = confirm; on = skip (sticky); once = skip next user turn then off.
+    yolo: YoloMode | None = None
     # Set by /edit; REPL sends as the next user turn then clears.
     pending_user_text: str | None = None
     client: ChatClient = field(init=False)
@@ -77,10 +80,32 @@ class ReplState:
             raise RuntimeError(msg)
         return key
 
-    def _confirm_destructive(self) -> bool:
-        if self.confirm_destructive is not None:
-            return self.confirm_destructive
-        return self.config.agent_config.confirm_destructive
+    def effective_yolo(self) -> YoloMode:
+        """Resolved YOLO mode (session override or config default)."""
+        if self.yolo is not None:
+            return self.yolo
+        return "off" if self.config.agent_config.confirm_destructive else "on"
+
+    def soft_confirm_enabled(self) -> bool:
+        """Whether destructive tools should prompt (or deny non-interactively)."""
+        return self.effective_yolo() == "off"
+
+    def set_yolo(self, mode: YoloMode) -> None:
+        """Set YOLO mode; rebuild tool registry when soft-confirm hooks change."""
+        prev = self.soft_confirm_enabled()
+        self.yolo = mode
+        if prev != self.soft_confirm_enabled():
+            self.rebuild_client()
+
+    def expire_yolo_once(self, *, quiet: bool = False) -> None:
+        """If mode is ``once``, drop back to ``off`` after a user turn."""
+        if self.effective_yolo() != "once":
+            return
+        self.set_yolo("off")
+        if not quiet:
+            import click
+
+            click.secho("yolo=off (once expired)", fg="bright_black", err=True)
 
     def _tool_registry(self) -> ToolRegistry | None:
         if not self.tools_enabled:
@@ -88,7 +113,7 @@ class ReplState:
         from plyngent.cli.limits import prompt_confirm_tool_async
         from plyngent.tools.danger import classify_danger
 
-        if self._confirm_destructive():
+        if self.soft_confirm_enabled():
             return ToolRegistry(
                 list(DEFAULT_TOOLS),
                 danger=classify_danger,
@@ -142,6 +167,13 @@ class ReplState:
         self._remote_models_key = None
         self._remote_models_error = None
 
+    def seed_remote_models(self, ids: list[str]) -> None:
+        """Install a freshly fetched remote catalog into the session cache."""
+        self._remote_models = list(ids)
+        self._remote_models_fetched_at = time.monotonic()
+        self._remote_models_key = self._models_cache_key()
+        self._remote_models_error = None
+
     def cached_remote_models(self) -> list[str] | None:
         """Return cached remote ids if still valid for the current provider."""
         if self._remote_models is None or self._remote_models_fetched_at is None:

src/plyngent/tools/chat/ask.py

@@ -4,9 +4,9 @@ from plyngent.agent import tool
 from plyngent.prompting import NonInteractiveError, ask_async
 
 
-@tool
+@tool(name="ask_user_line")
 async def ask_user(question: str, default: str = "") -> str:
-    """Ask the human a free-form question and return their answer.
+    """Ask the human a free-form one-line question and return their answer.
 
     Always allows arbitrary text. Use for clarifying requirements, preferences,
     or any input that is not a fixed menu. Optional ``default`` is used if the

src/plyngent/tools/chat/choose.py

@@ -46,7 +46,7 @@ def parse_options(raw: str) -> list[ChoiceOption]:
     return out
 
 
-@tool
+@tool(name="ask_user_choice")
 async def choose_user(
     question: str,
     options: str,
@@ -54,7 +54,7 @@ async def choose_user(
     *,
     allow_custom: bool = True,
 ) -> str:
-    """Ask the human to pick from options (or type a custom answer).
+    """Ask the human to pick from a list of options (or type a custom answer).
 
     ``options`` is a JSON array of strings, or objects with
     ``label``, optional ``description``, optional ``value``.

src/plyngent/tools/chat/form.py

@@ -54,13 +54,13 @@ def parse_fields(raw: str) -> list[FormField]:
     return out
 
 
-@tool
+@tool(name="ask_user_form")
 async def form_user(title: str, fields: str, *, confirm_submit: bool = True) -> str:
     """Run a multi-step form with the human; returns JSON object of answers.
 
     ``fields`` is a JSON array of objects:
     ``name``, ``prompt``, optional ``default``, optional ``options`` (same shape
-    as choose_user), optional ``allow_custom`` (default true).
+    as ask_user_choice), optional ``allow_custom`` (default true).
     When ``confirm_submit`` is true, the human reviews a summary before submit.
     """
     try:

tests/test_agent/test_confirm.py

@@ -56,4 +56,14 @@ async def test_confirm_deny() -> None:
 
 async def test_no_hooks_skips_confirm() -> None:
     registry = ToolRegistry([delete_path])
+    assert registry.soft_confirm is False
     assert await registry.execute("delete_path", '{"path": "a.txt"}') == "deleted a.txt"
+
+
+async def test_soft_confirm_property() -> None:
+    def on_confirm(name: str, args: Mapping[str, object], reason: str) -> bool:
+        del name, args, reason
+        return True
+
+    gated = ToolRegistry([delete_path], danger=classify_danger, on_confirm=on_confirm)
+    assert gated.soft_confirm is True

tests/test_cli/test_display.py

@@ -37,13 +37,47 @@ async def test_render_reasoning_and_text(capsys: pytest.CaptureFixture[str]) ->
         )
     )
     out = capsys.readouterr().out
-    assert "reasoning: " in out
+    assert "reasoning:" in out
     assert "think" in out
-    assert "assistant: " in out
+    assert "assistant:" in out
     assert "hello" in out
+    # Labels on their own lines (content begins after newline).
+    assert "assistant:\nhello" in out or "assistant:\r\nhello" in out
     set_markdown_enabled(True)
 
 
+async def test_flush_markdown_on_source_change(
+    capsys: pytest.CaptureFixture[str],
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    """Assistant segment before tools is flushed so later text is a new segment."""
+    monkeypatch.setattr("plyngent.cli.display.markdown_render_available", lambda: True)
+    set_markdown_enabled(True)
+    from plyngent.agent import ToolCallEvent
+    from plyngent.lmproto.openai_compatible.model import (
+        AssistantFunctionTool,
+        AssistantFunctionToolCall,
+    )
+
+    call = AssistantFunctionToolCall(
+        id="1",
+        function=AssistantFunctionTool(name="read_file", arguments="{}"),
+    )
+    await render_events(
+        _aiter(
+            [
+                TextDeltaEvent(content="before **tool**"),
+                ToolCallEvent(tool_call=call),
+                TextDeltaEvent(content="after"),
+            ]
+        ),
+        markdown=True,
+    )
+    out = capsys.readouterr().out
+    assert "[tool]" in out
+    assert "after" in out
+
+
 async def test_tool_result_preview_vs_verbose(capsys: pytest.CaptureFixture[str]) -> None:
     long = "x" * 200
     msg = ToolChatMessage(content=long, tool_call_id="1")

tests/test_cli/test_models_source.py

@@ -12,10 +12,14 @@ from plyngent.cli.models_source import (
 from plyngent.config.models import ModelConfig, OpenAICompatibleProvider
 
 
-def test_merge_model_choices_union() -> None:
-    assert merge_model_choices(["b", "a"], ["a", "c"]) == ["a", "b", "c"]
+def test_merge_model_choices_remote_first() -> None:
+    # remote-first: remote sorted, then config-only
+    assert merge_model_choices(["b", "a"], ["a", "c"]) == ["a", "c", "b"]
     assert merge_model_choices(["a"], None) == ["a"]
     assert merge_model_choices([], ["z"]) == ["z"]
+    assert merge_model_choices(["cfg"], ["remote", "cfg"], prefer="remote") == ["cfg", "remote"]
+    assert merge_model_choices(["b", "a"], ["a", "c"], prefer="union") == ["a", "b", "c"]
+    assert merge_model_choices(["b", "a"], ["a", "c"], prefer="config") == ["a", "b", "c"]
 
 
 def test_model_choices_for_provider() -> None:
@@ -26,6 +30,7 @@ def test_model_choices_for_provider() -> None:
     )
     assert config_model_ids(provider) == ["cfg"]
     assert model_choices_for_provider(provider, remote_ids=["remote", "cfg"]) == ["cfg", "remote"]
+    assert model_choices_for_provider(provider, remote_ids=["remote"]) == ["remote", "cfg"]
 
 
 def test_client_supports_models() -> None:

tests/test_cli/test_readline_setup.py

@@ -181,3 +181,5 @@ def test_complete_slash_args_from_registry(tmp_path: object) -> None:
     assert complete_slash_args(state, "/model", "a") == ["alpha"]
     assert complete_slash_args(state, "/export", "j") == ["json"]
     assert complete_slash_args(state, "/help", "st") == ["status", "stream"]
+    assert complete_slash_args(state, "/yolo", "") == ["on", "off", "once"]
+    assert complete_slash_args(state, "/yolo", "o") == ["on", "off", "once"]

tests/test_cli/test_repl_commands.py

@@ -341,6 +341,63 @@ async def test_verbose_toggle(state: ReplState) -> None:
     assert get_verbose_tool_results() is False
 
 
+async def test_yolo_toggle(state: ReplState, capsys: pytest.CaptureFixture[str]) -> None:
+    assert state.effective_yolo() == "off"
+    assert state.soft_confirm_enabled() is True
+    assert await handle_slash(state, "/yolo") is True
+    assert "yolo=off" in capsys.readouterr().out
+
+    assert await handle_slash(state, "/yolo on") is True
+    assert state.effective_yolo() == "on"
+    assert state.soft_confirm_enabled() is False
+    assert "yolo=on" in capsys.readouterr().out
+
+    assert await handle_slash(state, "/yolo once") is True
+    assert state.effective_yolo() == "once"
+    assert state.soft_confirm_enabled() is False
+
+    state.expire_yolo_once(quiet=True)
+    assert state.effective_yolo() == "off"
+    assert state.soft_confirm_enabled() is True
+
+    assert await handle_slash(state, "/yolo once") is True
+    state.expire_yolo_once()
+    err = capsys.readouterr().err
+    assert "yolo=off (once expired)" in err
+    assert state.effective_yolo() == "off"
+
+
+async def test_yolo_rebuilds_tool_registry(tmp_path: Path) -> None:
+    _ = set_workspace_root(tmp_path)
+    memory = await MemoryStore.open(DatabaseConfig())
+    provider = OpenAIProvider(access_key_or_token="sk-test")
+    config = ConfigStore(path=tmp_path / "plyngent.toml", document=tomlkit.document())
+    config.providers = {"local": provider}
+    st = ReplState(
+        config=config,
+        memory=memory,
+        workspace=tmp_path,
+        provider_name="local",
+        provider=provider,
+        model="gpt-test",
+        tools_enabled=True,
+    )
+    try:
+        assert st.agent.tools is not None
+        assert st.agent.tools.soft_confirm is True
+        st.set_yolo("on")
+        assert st.agent.tools is not None
+        assert st.agent.tools.soft_confirm is False
+        st.set_yolo("once")
+        assert st.agent.tools is not None
+        assert st.agent.tools.soft_confirm is False
+        st.set_yolo("off")
+        assert st.agent.tools is not None
+        assert st.agent.tools.soft_confirm is True
+    finally:
+        await memory.close()
+
+
 async def test_resume(state: ReplState) -> None:
     sid = state.session_id
     assert sid is not None
@@ -378,6 +435,7 @@ async def test_status_shows_context_tokens(state: ReplState, capsys: pytest.Capt
     assert "(est)" in out  # no API usage yet
     assert "context_chars=" in out
     assert "tool_result_max=" in out
+    assert "yolo=off" in out
     assert str(state.workspace) in out
 
     state.agent.last_request_usage = TokenUsage(

tests/test_cli/test_retry.py

@@ -5,7 +5,14 @@ from typing import TYPE_CHECKING, Literal, overload
 import pytest
 
 from plyngent.agent import ChatAgent
-from plyngent.cli.retry import retry_pending_with_retries, run_turn_with_retries, sleep_cancellable
+from plyngent.cli.retry import (
+    DEFAULT_MAX_AUTO_RETRIES,
+    DEFAULT_RETRY_DELAYS_SECONDS,
+    default_retry_delays,
+    retry_pending_with_retries,
+    run_turn_with_retries,
+    sleep_cancellable,
+)
 from plyngent.config.models import DatabaseConfig
 from plyngent.lmproto.openai_compatible.model import (
     AssistantChatMessage,
@@ -23,6 +30,17 @@ if TYPE_CHECKING:
     from collections.abc import AsyncIterator
 
 
+def test_default_retry_delays_schedule() -> None:
+    assert DEFAULT_MAX_AUTO_RETRIES == 10
+    assert default_retry_delays(0) == ()
+    assert default_retry_delays(4) == (5.0, 10.0, 15.0, 20.0)
+    delays = default_retry_delays(10)
+    assert delays[:4] == (5.0, 10.0, 15.0, 20.0)
+    assert delays[4:] == (30.0, 40.0, 50.0, 60.0, 70.0, 80.0)
+    assert delays == DEFAULT_RETRY_DELAYS_SECONDS
+    assert len(DEFAULT_RETRY_DELAYS_SECONDS) == 10
+
+
 def _response(message: AssistantChatMessage) -> ChatCompletionResponse:
     return ChatCompletionResponse(
         id="1",

tests/test_tools/test_chat_tools.py

@@ -12,7 +12,7 @@ async def test_ask_user_tool() -> None:
     backend = ScriptedBackend(["42"])
     with temporary_backend(backend):
         registry = ToolRegistry([ask_user])
-        out = await registry.execute("ask_user", '{"question": "Answer?"}')
+        out = await registry.execute("ask_user_line", '{"question": "Answer?"}')
     assert out == "42"
 
 
@@ -21,7 +21,7 @@ async def test_choose_user_tool_index() -> None:
     with temporary_backend(backend):
         registry = ToolRegistry([choose_user])
         out = await registry.execute(
-            "choose_user",
+            "ask_user_choice",
             json.dumps(
                 {
                     "question": "Pick",
@@ -36,7 +36,7 @@ async def test_choose_user_tool_index() -> None:
 async def test_choose_user_bad_options() -> None:
     registry = ToolRegistry([choose_user])
     out = await registry.execute(
-        "choose_user",
+        "ask_user_choice",
         json.dumps({"question": "Pick", "options": "not-json"}),
     )
     assert out.startswith("error:")
@@ -48,7 +48,7 @@ async def test_form_user_tool() -> None:
     with temporary_backend(backend):
         registry = ToolRegistry([form_user])
         out = await registry.execute(
-            "form_user",
+            "ask_user_form",
             json.dumps({"title": "Setup", "fields": fields, "confirm_submit": True}),
         )
     assert json.loads(out) == {"user": "ncbm"}
@@ -56,11 +56,11 @@ async def test_form_user_tool() -> None:
 
 async def test_chat_tools_in_default_list() -> None:
     names = {t.name for t in CHAT_TOOLS}
-    assert names == {"ask_user", "choose_user", "form_user"}
+    assert names == {"ask_user_line", "ask_user_choice", "ask_user_form"}
 
 
 async def test_ask_user_non_interactive_error() -> None:
     with temporary_backend(NonInteractiveBackend()):
         registry = ToolRegistry([ask_user])
-        out = await registry.execute("ask_user", '{"question": "hi"}')
+        out = await registry.execute("ask_user_line", '{"question": "hi"}')
     assert out.startswith("error:")