github copilot regularly prompts for passwords that can't be entered.

[ChefJayTay]

Type: Bug

I've had 3 occurances, just today, wasting considerable money, where this thing opens a terminal I can't access, performs an action requiring a sudo password, then halting with me having no way of entering said password. wasting entire prompt.

VS Code version: Code 1.127.0 (4fe60c8, 2026-06-30T10:52:33+02:00)
OS version: Linux x64 7.0.11-76070011-generic
Modes:

System Info

Item	Value
CPUs	AMD Ryzen 7 5800XT 8-Core Processor (16 x 3836)
GPU Status	2d_canvas: enabled GPU0: VENDOR= 0x10de, DEVICE=0x1f07, DRIVER_VENDOR=NVIDIA, DRIVER_VERSION=580.159.03 ACTIVE GPU1: VENDOR= 0x10de, DEVICE=0x2504 Machine model name: Machine model version: direct_rendering_display_compositor: disabled_off_ok gpu_compositing: enabled multiple_raster_threads: enabled_on opengl: enabled_on rasterization: enabled raw_draw: disabled_off_ok skia_graphite: disabled_off trees_in_viz: disabled_off video_decode: enabled video_encode: disabled_software vulkan: enabled_on webgl: enabled webgpu: enabled webgpu_on_vk_via_gl_interop: enabled webnn: disabled_off
Load (avg)	0, 1, 1
Memory (System)	31.25GB (21.55GB free)
Process Argv	--crash-reporter-id b045f875-8bae-47f6-82b1-2a85530b4f9a
Screen Reader	no
VM	0%
DESKTOP_SESSION	undefined
XDG_CURRENT_DESKTOP	COSMIC
XDG_SESSION_DESKTOP	COSMIC
XDG_SESSION_TYPE	wayland

Extensions (2)

Name	Identifier	Author	Version
Container Tools	ms-azuretools.vscode-containers	ms-azuretools	2.4.5
Dev Containers	ms-vscode-remote.remote-containers	ms-vscode-remote	0.463.0

A/B Experiments

vsliv368cf:30146710
vswsl492:30256859
binariesv615:30325510
nativeloc1:31344060
dwcopilot:31170013
dwoutputs:31242946
copilot_t_ci:31333650
e5gg6876:31282496
pythonrdcb7:31342333
6518g693:31463988
aj953862:31281341
63221493:31336333
envsactivate1:31538736
cloudbuttont:31379625
3efgi100_wstrepl:31403338
ec5jj548:31422691
cp_cls_t_966_ss:31526232
4je02754:31466945
8hhj4413:31478653
cp_cls_c_1081:31454833
conptydll_true:31498968
ia-use-proxy-models-svc:31452481
e9c30283:31461165
test_treatment2:31471001
c9b86496:31447327
idci7584:31464702
e3e4d672:31494082
ei9d7968:31496641
nes-extended-on:31455476
chat:31457767
8hig5102:31480529
89g7j272:31518289
7e187181:31503455
i2gc6536:31499202
52612955:31516516
ghj88844:31499326
ddid_c:31478207
hmra_i5g22:31518061
7df3h592:31512476
cp_cls_t_1082:31535311
logging_enabled_new:31498466
db5d2638:31499441
jb_cp_cls_c_632:31537245
b1ei0813:31506238
56dj4588:31512888
32d76977:31512328
ha629193:31508444
cdk-lw-on:31524445
2cfga817_c:31522397
jh5f2457_c:31539316
g4a0e106:31529594
41836166:31537719
pythondependencyremedydeactivate:31540010
jbcp_cls_pctr_t:31531130
61138546:31518536
cp_intellij_t_nes:31533504
cigf5339:31538644
jf4hg949:31526829
nes-curr-leftover-1500:31538366
h561g132:31533256
sessions_5:31532894
ihg5j128:31534457

[davemcom]

@ChefJayTay
They hide the terminal by default. It's challenging to convey how useful to agentic orchestration better terminal reuse strategy could be.

TLDR: Have you tried the experimental flag that's subject to removal?

"chat.tools.terminal.outputLocation":"terminal".

As outlined in response to comments found in this one Microsoft study where users were displeased with this feature.

• Sessions: Terminal and command output is not visible when the agent runs commands, showing "The terminal is no longer available" #302109 (comment)

The flag was interduced here:

• Test: terminal chat output location and actions #274926

[...] introduced chat.tools.terminal.outputLocation, which determines whether the chat created terminals are displayed.

Documented, but not documented.

The above load bearing word there is "whether" regarding terminals being displayed. This language and intent are NOT captured in the documentation of the experimental setting that keeps this escape hatch possible.

As of 7/4/2026, the intentional behavior of terminal retention is not clearly stated in the docs:

1. AI Settings: Agent Settings
2. Chat-Tools

Additional Lite reading:

• Easier way to clean up leftover terminals from chat sessions #290977
• Preserve user agency when cleaning up completed agent terminals #322466

[meganrogge]

It is a security issue to prompt for the password within the chat experience. What we do instead is tell the user that a password is required and in interactive mode, show a chat part that allows focusing the terminal.

In auto-approve/autopilot mode, we tell the user to run it interactively. Since we must imagine in these modes that a user is absent, waiting for further input before continuing would be bad and lead to hangs.

I think an improvement here would be to add an instruction like don't use sudo commands.

[meganrogge]

Note that the model might still run such commands