Gemini CLI 0.38.0

What's Changed

• fix(cli): refresh slash command list after /skills reload
• Update README.md for links.
• fix(core): ensure complete_task tool calls are recorded in chat history
• feat(policy): explicitly allow web_fetch in plan mode with ask_user
• fix(core): refactor linux sandbox to fix ARG_MAX crashes
• feat(config): add experimental.adk.agentSessionNoninteractiveEnabled setting
• Changelog for v0.36.0-preview.8
• feat(cli): change default loadingPhrases to 'off' to hide tips
• fix(cli): ensure agent stops when all declinable tools are cancelled
• fix(core): enhance sandbox usability and fix build error
• Terminal Serializer Optimization
• Auto configure memory.
• Unused error variables in catch block are not allowed
• feat(core): add background memory service for skill extraction
• feat: implement high-signal PR regression check for evaluations
• Fix shell output display
• fix(ui): resolve unwanted vertical spacing around various tool output treatments
• revert(cli): bring back input box and footer visibility in copy mode
• fix(cli): prevent crash in AnsiOutputText when handling non-array data
• feat(cli): support default values for environment variables
• Implement background process monitoring and inspection tools
• docs(browser-agent): update stale browser agent documentation
• fix: enable browser_agent in integration tests and add localhost fixture tests
• fix(browser): handle computer-use model detection for analyze_screenshot
• feat(core): Land ContextCompressionService
• feat(core): scope subagent workspace directories via AsyncLocalStorage
• Update ink version to 6.6.7
• fix(acp): handle all InvalidStreamError types gracefully in prompt
• Fix crash when vim editor is not found in PATH on Windows
• fix(core): move project memory dir under tmp directory
• Enable 'Other' option for yesno question type
• fix(cli): clear stale retry/loading state after cancellation (#21096)
• Changelog for v0.37.0-preview.0
• feat(core): implement context-aware persistent policy approvals
• docs: move agent disabling instructions and update remote agent status
• feat(cli): migrate nonInteractiveCli to LegacyAgentSession
• fix(core): unsafe type assertions in Core File System #19712
• fix(ui): hide model quota in /stats and refactor quota display
• Changelog for v0.36.0
• Changelog for v0.37.0-preview.1
• docs: add missing.md extensions to internal doc links
• fix(ui): fixed table styling
• fix(core): pass includeDirectories to sandbox configuration
• feat(ui): enable "TerminalBuffer" mode to solve flicker
• docs: clarify release coordination
• fix(core): remove broken PowerShell translation and fix native write in Windows sandbox
• Add instructions for how to start react in prod and force react to prod mode
• feat(cli): minimalist sandbox status labels
• Feat/browser agent metrics
• test: fix Windows CI execution and resolve exposed platform failures
• feat(core,cli): prioritize summary for topics (#24608)
• show color
• feat(cli): enable compact tool output by default (#24509)
• fix(core): inject skill system instructions into subagent prompts if activated
• fix(core): improve windows sandbox reliability and fix integration tests
• fix(core): ensure sandbox approvals are correctly persisted and matched for proactive expansions
• feat(cli) Scrollbar for input prompt
• Do not run pr-eval workflow when no steering changes detected
• Fix restoration of topic headers.
• feat(core): discourage update topic tool for simple tasks
• fix(core): ensure global temp directory is always in sandbox allowed paths
• fix(core): detect uninitialized lines
• docs: update sandboxing documentation and toolSandboxing settings
• feat(cli): enhance tool confirmation UI and selection layout
• feat(acp): add support for /about command
• feat(cli): add role specific metrics to /stats
• split context
• fix(cli): remove -S from shebang to fix Windows and BSD execution
• Fix issue where topic headers can be posted back to back
• fix(core): handle partial llm_request in BeforeModel hook override
• fix(ui): improve narration suppression and reduce flicker
• fix(ui): fixed auth race condition causing logo to flicker
• fix(browser): remove premature browser cleanup after subagent invocation
• Revert "feat(core,cli): prioritize summary for topics (#24608)"
• relax tool sandboxing overrides for plan mode to match defaults.
• fix(cli): respect global environment variable allowlist
• fix(cli): ensure skills list outputs to stdout in non-interactive environments
• Add an eval for and fix unsafe cloning behavior.
• fix(policy): allow complete_task in plan mode
• feat(telemetry): add browser agent clearcut metrics
• feat(cli): support selective topic expansion and click-to-expand
• temporarily disable sandbox integration test on windows
• Remove flakey test
• Alisa/approve button
• feat(hooks): display hook system messages in UI
• fix(core): propagate BeforeModel hook model override end-to-end
• chore: fix formatting for behavioral eval skill reference file
• fix: use directory junctions on Windows for skill linking
• fix(cli): prevent multiple banner increments on remount
• feat(acp): add /help command
• fix(core): remove tmux alternate buffer warning
• Improve sandbox error matching and caching
• feat(core): add agent protocol UI types and experimental flag
• feat(core): use experiment flags for default fetch timeouts
• Revert "fix(ui): improve narration suppression and reduce flicker (#2…
• refactor(cli): remove duplication in interactive shell awaiting input hint
• refactor(core): make LegacyAgentSession dependencies optional
• Changelog for v0.37.0-preview.2
• fix(cli): always show shell command description or actual command
• Added flag for ept size and increased default size
• fix(core): dispose Scheduler to prevent McpProgress listener leak
• fix(cli): switch default back to terminalBuffer=false and fix regressions introduced for that mode
• feat(cli): switch to ctrl+g from ctrl-x
• fix: isolate concurrent browser agent instances
• docs: update MCP server OAuth redirect port documentation