feat: add validated command quick actions by dakdevs · Pull Request #1 · dakdevs/t3code

dakdevs · 2026-08-10T02:06:12Z

What changed

detects executable quick actions only in fenced code blocks from completed assistant responses
validates shell syntax, executables, directories, and package scripts without invoking another model
runs the persisted command verbatim in a thread terminal without starting an agent turn
uses POSIX terminal job control to detect when a command actually blocks for terminal input
changes the inline result to Input required and opens the same live terminal session so the user can continue it
uses an explicit output-start boundary so shell line-editor redraws never leak into inline output
persists execution state and output per thread so leaving and reopening a thread, or viewing it from another client, restores the completed state and output
attaches completed output as hidden context only when the next user turn is sent
disables command quick actions on native Windows; WSL remains supported because its host is Linux

Why

Agents often give users commands to run manually. This adds a conservative, deterministic path for offering those commands as one-click actions while keeping execution explicit. Interactive commands are detected by the terminal itself when POSIX job control stops the process group for terminal access, so there is no command-name allowlist to maintain.

Prior art

pingdotgg/t3code#472 added a markdown run button, but was closed without merging and did not validate commands, persist execution state, detect interactivity, or carry output into the next turn.
pingdotgg/t3code#1032 added manual terminal-output context. This PR reuses the same product idea for automatic next-turn context after a quick action completes.

Validation

12 focused test files: 117 tests passed
real PTY test verifies SIGTTIN detection, resuming the same job, terminal input, output, and exit code
targeted lint and repository formatting passed
server, contracts, client-runtime, web, and mobile typechecks passed
mobile static checks passed; native linters were unavailable and no native files changed
diff check passed
the repository does not define a test:e2e script

Built with GPT-5 using the Codex desktop harness.

…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…dotgg#5964)

…tgg#5726)

…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-authored-by: codex <codex@users.noreply.github.com>

…g#4849) Co-authored-by: Julius Marminge <julius0216@outlook.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: maria <maria@kuuro.net> Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com> Co-authored-by: codex <codex@users.noreply.github.com>

…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

…gg#6013)

github-actions · 2026-08-11T20:09:35Z

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for 2ee38f7.

This comment will update automatically after the next completed run.

dakdevs · 2026-08-10T02:06:12Z

What changed

detects executable quick actions only in fenced code blocks from completed assistant responses
validates shell syntax, executables, directories, and package scripts without invoking another model
runs the persisted command verbatim in a thread terminal without starting an agent turn
uses POSIX terminal job control to detect when a command actually blocks for terminal input
changes the inline result to Input required and opens the same live terminal session so the user can continue it
uses an explicit output-start boundary so shell line-editor redraws never leak into inline output
persists execution state and output per thread so leaving and reopening a thread, or viewing it from another client, restores the completed state and output
attaches completed output as hidden context only when the next user turn is sent
disables command quick actions on native Windows; WSL remains supported because its host is Linux

Why

Agents often give users commands to run manually. This adds a conservative, deterministic path for offering those commands as one-click actions while keeping execution explicit. Interactive commands are detected by the terminal itself when POSIX job control stops the process group for terminal access, so there is no command-name allowlist to maintain.

Prior art

pingdotgg/t3code#472 added a markdown run button, but was closed without merging and did not validate commands, persist execution state, detect interactivity, or carry output into the next turn.
pingdotgg/t3code#1032 added manual terminal-output context. This PR reuses the same product idea for automatic next-turn context after a quick action completes.

Validation

12 focused test files: 117 tests passed
real PTY test verifies SIGTTIN detection, resuming the same job, terminal input, output, and exit code
targeted lint and repository formatting passed
server, contracts, client-runtime, web, and mobile typechecks passed
mobile static checks passed; native linters were unavailable and no native files changed
diff check passed
the repository does not define a test:e2e script

Built with GPT-5 using the Codex desktop harness.

…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…dotgg#5964)

…tgg#5726)

…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-authored-by: codex <codex@users.noreply.github.com>

…g#4849) Co-authored-by: Julius Marminge <julius0216@outlook.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: maria <maria@kuuro.net> Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com> Co-authored-by: codex <codex@users.noreply.github.com>

…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

…gg#6013)

github-actions · 2026-08-11T20:09:35Z

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for 2ee38f7.

This comment will update automatically after the next completed run.

dakdevs · 2026-08-10T02:06:12Z

What changed

detects executable quick actions only in fenced code blocks from completed assistant responses
validates shell syntax, executables, directories, and package scripts without invoking another model
runs the persisted command verbatim in a thread terminal without starting an agent turn
uses POSIX terminal job control to detect when a command actually blocks for terminal input
changes the inline result to Input required and opens the same live terminal session so the user can continue it
uses an explicit output-start boundary so shell line-editor redraws never leak into inline output
persists execution state and output per thread so leaving and reopening a thread, or viewing it from another client, restores the completed state and output
attaches completed output as hidden context only when the next user turn is sent
disables command quick actions on native Windows; WSL remains supported because its host is Linux

Why

Agents often give users commands to run manually. This adds a conservative, deterministic path for offering those commands as one-click actions while keeping execution explicit. Interactive commands are detected by the terminal itself when POSIX job control stops the process group for terminal access, so there is no command-name allowlist to maintain.

Prior art

pingdotgg/t3code#472 added a markdown run button, but was closed without merging and did not validate commands, persist execution state, detect interactivity, or carry output into the next turn.
pingdotgg/t3code#1032 added manual terminal-output context. This PR reuses the same product idea for automatic next-turn context after a quick action completes.

Validation

12 focused test files: 117 tests passed
real PTY test verifies SIGTTIN detection, resuming the same job, terminal input, output, and exit code
targeted lint and repository formatting passed
server, contracts, client-runtime, web, and mobile typechecks passed
mobile static checks passed; native linters were unavailable and no native files changed
diff check passed
the repository does not define a test:e2e script

Built with GPT-5 using the Codex desktop harness.

…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…dotgg#5964)

…tgg#5726)

…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-authored-by: codex <codex@users.noreply.github.com>

…g#4849) Co-authored-by: Julius Marminge <julius0216@outlook.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: maria <maria@kuuro.net> Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com> Co-authored-by: codex <codex@users.noreply.github.com>

…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

…gg#6013)

github-actions · 2026-08-11T20:09:35Z

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for 2ee38f7.

This comment will update automatically after the next completed run.

dakdevs · 2026-08-10T02:06:12Z

What changed

detects executable quick actions only in fenced code blocks from completed assistant responses
validates shell syntax, executables, directories, and package scripts without invoking another model
runs the persisted command verbatim in a thread terminal without starting an agent turn
uses POSIX terminal job control to detect when a command actually blocks for terminal input
changes the inline result to Input required and opens the same live terminal session so the user can continue it
uses an explicit output-start boundary so shell line-editor redraws never leak into inline output
persists execution state and output per thread so leaving and reopening a thread, or viewing it from another client, restores the completed state and output
attaches completed output as hidden context only when the next user turn is sent
disables command quick actions on native Windows; WSL remains supported because its host is Linux

Why

Agents often give users commands to run manually. This adds a conservative, deterministic path for offering those commands as one-click actions while keeping execution explicit. Interactive commands are detected by the terminal itself when POSIX job control stops the process group for terminal access, so there is no command-name allowlist to maintain.

Prior art

pingdotgg/t3code#472 added a markdown run button, but was closed without merging and did not validate commands, persist execution state, detect interactivity, or carry output into the next turn.
pingdotgg/t3code#1032 added manual terminal-output context. This PR reuses the same product idea for automatic next-turn context after a quick action completes.

Validation

12 focused test files: 117 tests passed
real PTY test verifies SIGTTIN detection, resuming the same job, terminal input, output, and exit code
targeted lint and repository formatting passed
server, contracts, client-runtime, web, and mobile typechecks passed
mobile static checks passed; native linters were unavailable and no native files changed
diff check passed
the repository does not define a test:e2e script

Built with GPT-5 using the Codex desktop harness.

…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…dotgg#5964)

…tgg#5726)

…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-authored-by: codex <codex@users.noreply.github.com>

…g#4849) Co-authored-by: Julius Marminge <julius0216@outlook.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: maria <maria@kuuro.net> Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com> Co-authored-by: codex <codex@users.noreply.github.com>

…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

…gg#6013)

github-actions · 2026-08-11T20:09:35Z

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for 2ee38f7.

This comment will update automatically after the next completed run.

dakdevs · 2026-08-10T02:06:12Z

What changed

detects executable quick actions only in fenced code blocks from completed assistant responses
validates shell syntax, executables, directories, and package scripts without invoking another model
runs the persisted command verbatim in a thread terminal without starting an agent turn
uses POSIX terminal job control to detect when a command actually blocks for terminal input
changes the inline result to Input required and opens the same live terminal session so the user can continue it
uses an explicit output-start boundary so shell line-editor redraws never leak into inline output
persists execution state and output per thread so leaving and reopening a thread, or viewing it from another client, restores the completed state and output
attaches completed output as hidden context only when the next user turn is sent
disables command quick actions on native Windows; WSL remains supported because its host is Linux

Why

Agents often give users commands to run manually. This adds a conservative, deterministic path for offering those commands as one-click actions while keeping execution explicit. Interactive commands are detected by the terminal itself when POSIX job control stops the process group for terminal access, so there is no command-name allowlist to maintain.

Prior art

pingdotgg/t3code#472 added a markdown run button, but was closed without merging and did not validate commands, persist execution state, detect interactivity, or carry output into the next turn.
pingdotgg/t3code#1032 added manual terminal-output context. This PR reuses the same product idea for automatic next-turn context after a quick action completes.

Validation

12 focused test files: 117 tests passed
real PTY test verifies SIGTTIN detection, resuming the same job, terminal input, output, and exit code
targeted lint and repository formatting passed
server, contracts, client-runtime, web, and mobile typechecks passed
mobile static checks passed; native linters were unavailable and no native files changed
diff check passed
the repository does not define a test:e2e script

Built with GPT-5 using the Codex desktop harness.

…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…dotgg#5964)

…tgg#5726)

…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-authored-by: codex <codex@users.noreply.github.com>

…g#4849) Co-authored-by: Julius Marminge <julius0216@outlook.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: maria <maria@kuuro.net> Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com> Co-authored-by: codex <codex@users.noreply.github.com>

…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

…gg#6013)

github-actions · 2026-08-11T20:09:35Z

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for 2ee38f7.

This comment will update automatically after the next completed run.

dakdevs · 2026-08-10T02:06:12Z

What changed

detects executable quick actions only in fenced code blocks from completed assistant responses
validates shell syntax, executables, directories, and package scripts without invoking another model
runs the persisted command verbatim in a thread terminal without starting an agent turn
uses POSIX terminal job control to detect when a command actually blocks for terminal input
changes the inline result to Input required and opens the same live terminal session so the user can continue it
uses an explicit output-start boundary so shell line-editor redraws never leak into inline output
persists execution state and output per thread so leaving and reopening a thread, or viewing it from another client, restores the completed state and output
attaches completed output as hidden context only when the next user turn is sent
disables command quick actions on native Windows; WSL remains supported because its host is Linux

Why

Agents often give users commands to run manually. This adds a conservative, deterministic path for offering those commands as one-click actions while keeping execution explicit. Interactive commands are detected by the terminal itself when POSIX job control stops the process group for terminal access, so there is no command-name allowlist to maintain.

Prior art

pingdotgg/t3code#472 added a markdown run button, but was closed without merging and did not validate commands, persist execution state, detect interactivity, or carry output into the next turn.
pingdotgg/t3code#1032 added manual terminal-output context. This PR reuses the same product idea for automatic next-turn context after a quick action completes.

Validation

12 focused test files: 117 tests passed
real PTY test verifies SIGTTIN detection, resuming the same job, terminal input, output, and exit code
targeted lint and repository formatting passed
server, contracts, client-runtime, web, and mobile typechecks passed
mobile static checks passed; native linters were unavailable and no native files changed
diff check passed
the repository does not define a test:e2e script

Built with GPT-5 using the Codex desktop harness.

…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…dotgg#5964)

…tgg#5726)

…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-authored-by: codex <codex@users.noreply.github.com>

…g#4849) Co-authored-by: Julius Marminge <julius0216@outlook.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: maria <maria@kuuro.net> Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com> Co-authored-by: codex <codex@users.noreply.github.com>

…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

…gg#6013)

github-actions · 2026-08-11T20:09:35Z

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for 2ee38f7.

This comment will update automatically after the next completed run.

dakdevs · 2026-08-10T02:06:12Z

What changed

detects executable quick actions only in fenced code blocks from completed assistant responses
validates shell syntax, executables, directories, and package scripts without invoking another model
runs the persisted command verbatim in a thread terminal without starting an agent turn
uses POSIX terminal job control to detect when a command actually blocks for terminal input
changes the inline result to Input required and opens the same live terminal session so the user can continue it
uses an explicit output-start boundary so shell line-editor redraws never leak into inline output
persists execution state and output per thread so leaving and reopening a thread, or viewing it from another client, restores the completed state and output
attaches completed output as hidden context only when the next user turn is sent
disables command quick actions on native Windows; WSL remains supported because its host is Linux

Why

Agents often give users commands to run manually. This adds a conservative, deterministic path for offering those commands as one-click actions while keeping execution explicit. Interactive commands are detected by the terminal itself when POSIX job control stops the process group for terminal access, so there is no command-name allowlist to maintain.

Prior art

pingdotgg/t3code#472 added a markdown run button, but was closed without merging and did not validate commands, persist execution state, detect interactivity, or carry output into the next turn.
pingdotgg/t3code#1032 added manual terminal-output context. This PR reuses the same product idea for automatic next-turn context after a quick action completes.

Validation

12 focused test files: 117 tests passed
real PTY test verifies SIGTTIN detection, resuming the same job, terminal input, output, and exit code
targeted lint and repository formatting passed
server, contracts, client-runtime, web, and mobile typechecks passed
mobile static checks passed; native linters were unavailable and no native files changed
diff check passed
the repository does not define a test:e2e script

Built with GPT-5 using the Codex desktop harness.

…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…dotgg#5964)

…tgg#5726)

…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-authored-by: codex <codex@users.noreply.github.com>

…g#4849) Co-authored-by: Julius Marminge <julius0216@outlook.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: maria <maria@kuuro.net> Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com> Co-authored-by: codex <codex@users.noreply.github.com>

…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

…gg#6013)

github-actions · 2026-08-11T20:09:35Z

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for 2ee38f7.

This comment will update automatically after the next completed run.

dakdevs · 2026-08-10T02:06:12Z

What changed

detects executable quick actions only in fenced code blocks from completed assistant responses
validates shell syntax, executables, directories, and package scripts without invoking another model
runs the persisted command verbatim in a thread terminal without starting an agent turn
uses POSIX terminal job control to detect when a command actually blocks for terminal input
changes the inline result to Input required and opens the same live terminal session so the user can continue it
uses an explicit output-start boundary so shell line-editor redraws never leak into inline output
persists execution state and output per thread so leaving and reopening a thread, or viewing it from another client, restores the completed state and output
attaches completed output as hidden context only when the next user turn is sent
disables command quick actions on native Windows; WSL remains supported because its host is Linux

Why

Agents often give users commands to run manually. This adds a conservative, deterministic path for offering those commands as one-click actions while keeping execution explicit. Interactive commands are detected by the terminal itself when POSIX job control stops the process group for terminal access, so there is no command-name allowlist to maintain.

Prior art

pingdotgg/t3code#472 added a markdown run button, but was closed without merging and did not validate commands, persist execution state, detect interactivity, or carry output into the next turn.
pingdotgg/t3code#1032 added manual terminal-output context. This PR reuses the same product idea for automatic next-turn context after a quick action completes.

Validation

12 focused test files: 117 tests passed
real PTY test verifies SIGTTIN detection, resuming the same job, terminal input, output, and exit code
targeted lint and repository formatting passed
server, contracts, client-runtime, web, and mobile typechecks passed
mobile static checks passed; native linters were unavailable and no native files changed
diff check passed
the repository does not define a test:e2e script

Built with GPT-5 using the Codex desktop harness.

…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…dotgg#5964)

…tgg#5726)

…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-authored-by: codex <codex@users.noreply.github.com>

…g#4849) Co-authored-by: Julius Marminge <julius0216@outlook.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: maria <maria@kuuro.net> Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com> Co-authored-by: codex <codex@users.noreply.github.com>

…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

…gg#6013)

github-actions · 2026-08-11T20:09:35Z

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for 2ee38f7.

This comment will update automatically after the next completed run.

juliusmarmingeand others added 30 commits August 10, 2026 04:11

Automate production mobile EAS releases (pingdotgg#5609)
73b2e8f

Add settings and usage breadcrumbs (pingdotgg#5929)
9690680

fix(web): correct model picker trigger padding (pingdotgg#5935)
cbd55d6

fix(web): show worktree icon in sidebar v2 (pingdotgg#5909)
0ca9fb3

fix(web): enable restore defaults after theme mix changes (pingdotgg#…
ef051bd
…5928) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>

fix(web): trait menu closes after you pick a level (pingdotgg#5879)
d432100
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

fix(web): align project name with headline (pingdotgg#5864)
f0e2975

fix(web): update pills use readable theme foregrounds (pingdotgg#5938)
3d74474

fix(web): use themed confirmation dialogs (pingdotgg#5624)
9821bca

fix(web): use import/export-appropriate icons for theme buttons (ping…
fbcae59
…dotgg#5964)

fix(mobile): detect PowerShell cmdlet errors in work log rows (pingdo…
67ef189
…tgg#5726)

fix(mobile): stop Android user bubbles with code blocks from overlapp…
d440442
…ing (pingdotgg#5659) Co-authored-by: Rodrigo Brechard <rodrigo@clubtidy.fr> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

fix(mobile): parse EAS fingerprint JSON (pingdotgg#5991)
3b72d17
Co-authored-by: codex <codex@users.noreply.github.com>

chore(release): prepare v0.0.33
78f462c

perf(desktop): probe the Windows shell environment concurrently (ping…
8de0aa2
…dotgg#5878) Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>

feat(web): add duplicate action to the T3 Code default theme (pingdot…
b91a000
…gg#6013)

fix(web): improve built-in theme contrast (pingdotgg#6000)
0a7c662

fix(ci): extend release publish timeout (pingdotgg#6034)
1b120f3

Allow Android tablets to rotate (pingdotgg#5613)
428d9f9

feat: add validated command quick actions
e3c65dc

fix: apply live command quick actions
658a20c

feat: show quick action output inline
6d96f27

fix: keep quick actions inline
217f87c

fix: render terminal backspaces inline
abb0569

fix: trim idle shell prompts inline
b66aa0d

feat: run quick actions from code blocks
e23bd30

fix: make code block actions clickable
8a3570a

fix: preserve inline quick action output
1a99210

fix: stabilize quick action completion state
78a10d4

dakdevs added 5 commits August 10, 2026 12:36

fix: track quick action command completion
e721bd8

fix: show quick action exit codes
8b217cd

fix: hide quick action wrapper fragments
c71fbfc

fix: persist quick action executions per thread
40be99f

feat: continue interactive quick actions in terminal
022f539

dakdevs force-pushed the codex/command-quick-actions branch from 0dcaab1 to 022f539Compare August 10, 2026 19:39

github-actionsBot added vouch:trusted PR author is trusted by repo permissions or the VOUCHED list. size:XXL labels Aug 10, 2026

dakdevs added 4 commits August 10, 2026 12:52

fix: hide quick action shell redraws
6a8ea93

fix: clean persisted quick action redraws
795b185

fix: persist quick actions across navigation
a565038

fix: stream quick action state updates
2ee38f7