[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M74FQTQRMT45RBD50D71K8F3":3,"skill-same:01M74FQTQRMT45RBD50D71K8F3":44},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M74FQTQRMT45RBD50D71K8F3","agent-workspace-linux",null,"Use when a task needs a hidden Linux desktop running in the background or a workspace-owned browser: GUI app QA, web\u002Fbrowser\u002Fshopping automation, app observation, or stale workspace cleanup. Routes agent-workspace-linux MCP tools on demand. Does NOT apply to host desktop\u002FChrome control, generic MCP setup, or pure code\u002Ffile edits.","distsystem\u002Fcodex-desktop-linux","distsystem","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F233524749?v=4","01M74FQKS438KRTX3JXSNS3Q15","01M74FQKSN4TTD8W3M0ARTGP4W",17,"MIT","https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux\u002Fblob\u002Fmain\u002Flinux-features\u002Fagent-workspace\u002Fskills\u002Fagent-workspace-linux\u002FSKILL.md",[],[],"2026-11-11 15:11:14",false,"---\nname: agent-workspace-linux\ndescription: \"Use when a task needs a hidden Linux desktop running in the background or a workspace-owned browser: GUI app QA, web\u002Fbrowser\u002Fshopping automation, app observation, or stale workspace cleanup. Routes agent-workspace-linux MCP tools on demand. Does NOT apply to host desktop\u002FChrome control, generic MCP setup, or pure code\u002Ffile edits.\"\n---\n\n# agent-workspace-linux\n\nDrive an agent-owned Linux desktop and workspace-owned browser in the\nbackground through the `agent-workspace-linux` MCP server. The workspace runs\napps, browser automation, screenshots, clipboard, and input inside its own\nhidden display, never the user's real desktop, focus, or host Chrome.\n\nIt is not a sandbox. Apps you launch there run as the user, with the user's\nfiles, network, keyring and SSH agent: a `git push`, a sent message or a file\ndeleted from a workspace terminal is real. Act as carefully as on the user's\nown machine. Launch Chrome through `workspace_open_browser` or with `--user-data-dir`: a plain launch would hand off to the user's running browser and is refused.\n\nThis skill is the lightweight entry point. The MCP exposes many tools; do not\nload, list, or paste them all up front. Route by phase and ask the host to load\nonly the tool schemas needed for the next action.\n\n## Critical Boundaries\n\n### Use the dedicated Codex for Linux feature page for setup\nCodex for Linux owns Agent Workspace setup through the **Agent Workspaces** page:\nbinary path, page-authored permission rules, permission file mutation,\nReconnect\u002FSmoke test, profile creation, workspace start\u002Fstop, and viewer launch.\nDo not send users to generic MCP settings, general configuration pages, or\n`~\u002F.codex\u002Fconfig.toml` for normal Codex for Linux setup.\n\nSkip unless: the host is Codex for Linux and the task is setup, permission\nchanges, reconnect\u002Frestart, or the workspace tools are unavailable.\n\n### Keep tools progressive\nThe bundled installer is skill-first. `.\u002Finstall.sh` installs this skill under\n`~\u002F.codex\u002Fskills` by default and leaves generic Codex MCP config untouched\nunless the user explicitly chooses `--codex-configure` or `--permissions`.\nFor migration only, `.\u002Finstall.sh --clean-codex-config` removes stale generic\nCodex MCP entries left by older installs; restart or reconnect Codex afterward.\n\nSkip unless: the user asks how to install\u002Fregister the backend or why the tool\nfamily should not be loaded at startup.\n\n### Never target the host desktop\nUse this skill only for the hidden workspace. For the user's real Linux desktop,\nreal Chrome profile, existing tabs, or host focus\u002Fmouse\u002Fkeyboard, use the\nappropriate host-desktop or browser tool instead.\n\nSkip unless: the requested action can run inside the workspace display\nor workspace-owned browser.\n\n## When To Use\n\n- GUI app QA in a throwaway desktop.\n- Browser\u002Fweb\u002Fshopping automation that must not hijack host Chrome.\n- Observation of workspace windows, logs, events, screenshots, or artifacts.\n- Cleanup of stale or orphaned agent workspaces.\n\nSkip unless: the user task requires running or driving a real GUI app\u002Fbrowser\noff the user's screen. For pure code, shell, or file edits, do not start a workspace.\n\n## Permission Model\n\n- **Default**: no MCP ceiling is configured; the host\u002Fclient owns approvals.\n  Call `mcp_permissions` once before mutating to confirm `configured=false`.\n- **App allowlist**: `--permissions PATH` or `AGENT_WORKSPACE_PERMISSIONS`\n  limits which programs agents may launch, for the life of that MCP process. It can only be broadened by changing the config and\n  restarting\u002Freconnecting the backend from the owning setup surface.\n- **Action marks**: `mcp_action_catalog` marks each tool read-only, write,\n  destructive or open-world. Ask the user before destructive or open-world\n  actions they did not request.\n- **Live control**: `mcp_control_state` read-only\u002Fpaused\u002Factive is a\n  best-effort runtime control, not the authoritative boundary.\n- **Real-world actions**: checkout, purchases, account changes, and messages\n  require explicit user approval. Keep drafting separate from final action.\n\nSkip unless: you are about to start a workspace, open a profile, launch\u002Frun an\napp, send input, or act on an external account\u002Fsite.\n\n## Phase Router\n\nAlways orient before mutating.\n\n| Phase | Load only these tools |\n| --- | --- |\n| Orient | `mcp_agent_context`, `mcp_session_brief`, `mcp_task_plan`, `mcp_permissions`, `mcp_action_catalog`, `workspace_doctor`, `workspace_list`, `workspace_status` |\n| Profiles | `profile_list`, `profile_get`, `profile_template`, `profile_put`, `profile_check` |\n| Start | `workspace_start`, `workspace_open_profile` |\n| Observe | `workspace_observe`, `workspace_screenshot`, `workspace_list_windows`, `workspace_active_window`, `workspace_read_app_log`, `workspace_events` |\n| Act | `workspace_launch_app`, `workspace_run_app`, `workspace_click`, `workspace_type_text`, `workspace_key`, `workspace_paste_text`, `workspace_focus_window`, `workspace_close_window` |\n| Browser | `workspace_open_browser`, `workspace_browser_targets`, `workspace_browser_snapshot`, `workspace_browser_navigate`, `workspace_browser_search_results`, `workspace_browser_click` |\n| Viewer\u002Fcontrol | `mcp_control_state`, `mcp_control_update`, `workspace_open_viewer`, `workspace_list_viewers`, `workspace_close_viewer` |\n| Teardown | `workspace_stop`, `workspace_kill_app`, `workspace_cleanup_stale` |\n\nSkip unless: the current phase needs one of the listed capabilities. Do not load\nschemas for later phases until the plan reaches them.\n\n## Safe Workflow\n\n1. Orient: call `mcp_agent_context` or `mcp_session_brief`, then use\n   `mcp_task_plan` for app-QA, browser, observe, or cleanup intent.\n2. Check permissions: call `mcp_permissions` before mutating.\n3. Start: use `workspace_start` with `ack_hidden_workspace=true` and a clear\n   `purpose`, or `workspace_open_profile` for a saved profile.\n4. Observe: use `workspace_observe` or focused screenshot\u002Fwindow\u002Flog\u002Fevent\n   tools before sending input.\n5. Act: launch apps, run commands, click\u002Ftype\u002Fpaste\u002Fkey only inside the\n   workspace.\n6. Stop: use `workspace_stop` when finished; use `workspace_cleanup_stale` for\n   orphaned runtimes.\n\nSession selection supports X11 (`Xvfb` + Openbox), i3 on private Xvfb, and\nnative Wayland sessions on KWin, GNOME, Hyprland and COSMIC. MCP callers pass\n`session: {\"backend\":\"\u003Cname>\"}` with one of `x11`, `kwin`, `gnome`, `i3`,\n`hyprland` or `cosmic`, and may add `\"scale\":1.5` on the native Wayland\nbackends for 150% output scaling. CLI start\u002Fopen-profile accepts the same\nnames with `--session NAME --scale 1.5`. Explicit MCP input overrides the\nprofile session; omitting both keeps the pinned X11 default.\nKWin needs version 6.6 or newer, kwin-common, kscreen-doctor, wl-clipboard,\nand an accessible DRM render node. It runs directly, without a parent Xvfb, and\nis not a full Plasma shell. GNOME runs GNOME Shell 50 or 51 with a private\nheadless output and software rendering; Mutter accepts only scales the chosen\noutput mode supports. Hyprland uses a private KWin renderer and supports\nfractional output scales; see [the Hyprland backend guide](https:\u002F\u002Fgithub.com\u002Fagent-sh\u002Fagent-workspace-linux\u002Fblob\u002Fmain\u002Fdocs\u002Fhyprland.md).\nCOSMIC runs `cosmic-comp` under a software-only private Weston parent; see\n[the COSMIC guide](https:\u002F\u002Fgithub.com\u002Fagent-sh\u002Fagent-workspace-linux\u002Fblob\u002Fmain\u002Fdocs\u002Fnative-cosmic.md) for its compositor and\nXWayland dependencies.\nSelect i3 with `session: {\"backend\":\"i3\"}` or `--session i3`. It uses scale 1,\nneeds `i3-wm` and the X11 dependencies including ImageMagick `import`, and has\nno launcher bindings or startup commands. Move and resize float the target window.\nMinimize uses the scratchpad; show restores the window to the current workspace.\nPointer and window coordinates use screenshot pixels on every backend.\nUnsupported session names are rejected.\n\nSkip unless: each step's precondition is visible in the previous tool result\n(workspace id, app id, target window, browser target, or explicit user approval).\n\n## Browser Tasks\n\nUse the workspace-owned browser over its loopback DevTools endpoint. Start with\n`workspace_open_browser`, discover pages with `workspace_browser_targets`, read\nwith `workspace_browser_snapshot` or `workspace_browser_search_results`, then\nnavigate or click with browser tools.\n\nBrowser tools require a Chrome launched with `--user-data-dir` and loopback\nDevTools. Always get it from `workspace_open_browser` (MCP) or\n`workspace open-browser` (CLI) rather than hand-rolling a `workspace_launch_app`\nargv.\n\nIf you attach your own CDP\u002FWebSocket client to that endpoint, omit the `Origin`\nheader (e.g. `websocket-client`'s `suppress_origin=True`). The DevTools port is\nephemeral, so Chrome's origin allow-list cannot be pre-seeded and an `Origin`-\nbearing handshake is rejected with `403 Forbidden`.\n\nAfter `workspace_browser_navigate` or `workspace_browser_click`, read\n`page.url` \u002F `page.title` for post-action state. `target.*` is refreshed to match,\nbut if the refresh fails the response carries a warning saying it may be stale:\ncheck `warnings` before trusting `target` for verification.\n\nSkip unless: the task is web\u002Fbrowser automation that can run in the\nworkspace. Do not attach to host Chrome, Playwright, or Computer Use as a\nshortcut.\n\n## Do NOT\n\n- Do not dump all MCP tools into the agent context at startup.\n- Do not send users to generic MCP\u002Fconfiguration pages for Codex for Linux\n  Agent Workspace setup.\n- Do not start a workspace without `ack_hidden_workspace=true` and a `purpose`.\n- Do not send input to or screenshot the user's real desktop.\n- Do not perform purchases, checkout, account changes, or sends without explicit\n  user approval.\n- Do not leave workspaces running after the task.\n\n## Example Trajectory\n\nTask: \"QA the settings dialog of myapp.\"\n\n1. `mcp_agent_context` and `mcp_task_plan` for app-QA.\n2. `mcp_permissions` to confirm the active boundary.\n3. `workspace_start` with `ack_hidden_workspace=true` and purpose\n   `\"QA myapp settings\"`.\n4. `workspace_launch_app` with the app command.\n5. `workspace_observe` to find the window.\n6. `workspace_click` \u002F `workspace_type_text` to exercise Settings.\n7. `workspace_screenshot` and `workspace_read_app_log` for evidence.\n8. `workspace_stop`.\n\n## Constraints\n\n- Keep activation low-noise: only use this skill for background GUI\u002Fbrowser work.\n- Keep context small: frontmatter loads by default; body loads on activation;\n  tool schemas load on demand.\n- Omit `allowed-tools` intentionally: different hosts namespace MCP tools\n  differently, and the skill routes the full family progressively.\n- Validate with `agnix --target generic|claude-code|cursor|codex|kiro`\n  when changing this file.\n","# agent-workspace-linux\n\nDrive an agent-owned Linux desktop and workspace-owned browser in the\nbackground through the `agent-workspace-linux` MCP server. The workspace runs\napps, browser automation, screenshots, clipboard, and input inside its own\nhidden display, never the user's real desktop, focus, or host Chrome.\n\nIt is not a sandbox. Apps you launch there run as the user, with the user's\nfiles, network, keyring and SSH agent: a `git push`, a sent message or a file\ndeleted from a workspace terminal is real. Act as carefully as on the user's\nown machine. Launch Chrome through `workspace_open_browser` or with `--user-data-dir`: a plain launch would hand off to the user's running browser and is refused.\n\nThis skill is the lightweight entry point. The MCP exposes many tools; do not\nload, list, or paste them all up front. Route by phase and ask the host to load\nonly the tool schemas needed for the next action.\n\n## Critical Boundaries\n\n### Use the dedicated Codex for Linux feature page for setup\nCodex for Linux owns Agent Workspace setup through the **Agent Workspaces** page:\nbinary path, page-authored permission rules, permission file mutation,\nReconnect\u002FSmoke test, profile creation, workspace start\u002Fstop, and viewer launch.\nDo not send users to generic MCP settings, general configuration pages, or\n`~\u002F.codex\u002Fconfig.toml` for normal Codex for Linux setup.\n\nSkip unless: the host is Codex for Linux and the task is setup, permission\nchanges, reconnect\u002Frestart, or the workspace tools are unavailable.\n\n### Keep tools progressive\nThe bundled installer is skill-first. `.\u002Finstall.sh` installs this skill under\n`~\u002F.codex\u002Fskills` by default and leaves generic Codex MCP config untouched\nunless the user explicitly chooses `--codex-configure` or `--permissions`.\nFor migration only, `.\u002Finstall.sh --clean-codex-config` removes stale generic\nCodex MCP entries left by older installs; restart or reconnect Codex afterward.\n\nSkip unless: the user asks how to install\u002Fregister the backend or why the tool\nfamily should not be loaded at startup.\n\n### Never target the host desktop\nUse this skill only for the hidden workspace. For the user's real Linux desktop,\nreal Chrome profile, existing tabs, or host focus\u002Fmouse\u002Fkeyboard, use the\nappropriate host-desktop or browser tool instead.\n\nSkip unless: the requested action can run inside the workspace display\nor workspace-owned browser.\n\n## When To Use\n\n- GUI app QA in a throwaway desktop.\n- Browser\u002Fweb\u002Fshopping automation that must not hijack host Chrome.\n- Observation of workspace windows, logs, events, screenshots, or artifacts.\n- Cleanup of stale or orphaned agent workspaces.\n\nSkip unless: the user task requires running or driving a real GUI app\u002Fbrowser\noff the user's screen. For pure code, shell, or file edits, do not start a workspace.\n\n## Permission Model\n\n- **Default**: no MCP ceiling is configured; the host\u002Fclient owns approvals.\n  Call `mcp_permissions` once before mutating to confirm `configured=false`.\n- **App allowlist**: `--permissions PATH` or `AGENT_WORKSPACE_PERMISSIONS`\n  limits which programs agents may launch, for the life of that MCP process. It can only be broadened by changing the config and\n  restarting\u002Freconnecting the backend from the owning setup surface.\n- **Action marks**: `mcp_action_catalog` marks each tool read-only, write,\n  destructive or open-world. Ask the user before destructive or open-world\n  actions they did not request.\n- **Live control**: `mcp_control_state` read-only\u002Fpaused\u002Factive is a\n  best-effort runtime control, not the authoritative boundary.\n- **Real-world actions**: checkout, purchases, account changes, and messages\n  require explicit user approval. Keep drafting separate from final action.\n\nSkip unless: you are about to start a workspace, open a profile, launch\u002Frun an\napp, send input, or act on an external account\u002Fsite.\n\n## Phase Router\n\nAlways orient before mutating.\n\n| Phase | Load only these tools |\n| --- | --- |\n| Orient | `mcp_agent_context`, `mcp_session_brief`, `mcp_task_plan`, `mcp_permissions`, `mcp_action_catalog`, `workspace_doctor`, `workspace_list`, `workspace_status` |\n| Profiles | `profile_list`, `profile_get`, `profile_template`, `profile_put`, `profile_check` |\n| Start | `workspace_start`, `workspace_open_profile` |\n| Observe | `workspace_observe`, `workspace_screenshot`, `workspace_list_windows`, `workspace_active_window`, `workspace_read_app_log`, `workspace_events` |\n| Act | `workspace_launch_app`, `workspace_run_app`, `workspace_click`, `workspace_type_text`, `workspace_key`, `workspace_paste_text`, `workspace_focus_window`, `workspace_close_window` |\n| Browser | `workspace_open_browser`, `workspace_browser_targets`, `workspace_browser_snapshot`, `workspace_browser_navigate`, `workspace_browser_search_results`, `workspace_browser_click` |\n| Viewer\u002Fcontrol | `mcp_control_state`, `mcp_control_update`, `workspace_open_viewer`, `workspace_list_viewers`, `workspace_close_viewer` |\n| Teardown | `workspace_stop`, `workspace_kill_app`, `workspace_cleanup_stale` |\n\nSkip unless: the current phase needs one of the listed capabilities. Do not load\nschemas for later phases until the plan reaches them.\n\n## Safe Workflow\n\n1. Orient: call `mcp_agent_context` or `mcp_session_brief`, then use\n   `mcp_task_plan` for app-QA, browser, observe, or cleanup intent.\n2. Check permissions: call `mcp_permissions` before mutating.\n3. Start: use `workspace_start` with `ack_hidden_workspace=true` and a clear\n   `purpose`, or `workspace_open_profile` for a saved profile.\n4. Observe: use `workspace_observe` or focused screenshot\u002Fwindow\u002Flog\u002Fevent\n   tools before sending input.\n5. Act: launch apps, run commands, click\u002Ftype\u002Fpaste\u002Fkey only inside the\n   workspace.\n6. Stop: use `workspace_stop` when finished; use `workspace_cleanup_stale` for\n   orphaned runtimes.\n\nSession selection supports X11 (`Xvfb` + Openbox), i3 on private Xvfb, and\nnative Wayland sessions on KWin, GNOME, Hyprland and COSMIC. MCP callers pass\n`session: {\"backend\":\"\u003Cname>\"}` with one of `x11`, `kwin`, `gnome`, `i3`,\n`hyprland` or `cosmic`, and may add `\"scale\":1.5` on the native Wayland\nbackends for 150% output scaling. CLI start\u002Fopen-profile accepts the same\nnames with `--session NAME --scale 1.5`. Explicit MCP input overrides the\nprofile session; omitting both keeps the pinned X11 default.\nKWin needs version 6.6 or newer, kwin-common, kscreen-doctor, wl-clipboard,\nand an accessible DRM render node. It runs directly, without a parent Xvfb, and\nis not a full Plasma shell. GNOME runs GNOME Shell 50 or 51 with a private\nheadless output and software rendering; Mutter accepts only scales the chosen\noutput mode supports. Hyprland uses a private KWin renderer and supports\nfractional output scales; see [the Hyprland backend guide](https:\u002F\u002Fgithub.com\u002Fagent-sh\u002Fagent-workspace-linux\u002Fblob\u002Fmain\u002Fdocs\u002Fhyprland.md).\nCOSMIC runs `cosmic-comp` under a software-only private Weston parent; see\n[the COSMIC guide](https:\u002F\u002Fgithub.com\u002Fagent-sh\u002Fagent-workspace-linux\u002Fblob\u002Fmain\u002Fdocs\u002Fnative-cosmic.md) for its compositor and\nXWayland dependencies.\nSelect i3 with `session: {\"backend\":\"i3\"}` or `--session i3`. It uses scale 1,\nneeds `i3-wm` and the X11 dependencies including ImageMagick `import`, and has\nno launcher bindings or startup commands. Move and resize float the target window.\nMinimize uses the scratchpad; show restores the window to the current workspace.\nPointer and window coordinates use screenshot pixels on every backend.\nUnsupported session names are rejected.\n\nSkip unless: each step's precondition is visible in the previous tool result\n(workspace id, app id, target window, browser target, or explicit user approval).\n\n## Browser Tasks\n\nUse the workspace-owned browser over its loopback DevTools endpoint. Start with\n`workspace_open_browser`, discover pages with `workspace_browser_targets`, read\nwith `workspace_browser_snapshot` or `workspace_browser_search_results`, then\nnavigate or click with browser tools.\n\nBrowser tools require a Chrome launched with `--user-data-dir` and loopback\nDevTools. Always get it from `workspace_open_browser` (MCP) or\n`workspace open-browser` (CLI) rather than hand-rolling a `workspace_launch_app`\nargv.\n\nIf you attach your own CDP\u002FWebSocket client to that endpoint, omit the `Origin`\nheader (e.g. `websocket-client`'s `suppress_origin=True`). The DevTools port is\nephemeral, so Chrome's origin allow-list cannot be pre-seeded and an `Origin`-\nbearing handshake is rejected with `403 Forbidden`.\n\nAfter `workspace_browser_navigate` or `workspace_browser_click`, read\n`page.url` \u002F `page.title` for post-action state. `target.*` is refreshed to match,\nbut if the refresh fails the response carries a warning saying it may be stale:\ncheck `warnings` before trusting `target` for verification.\n\nSkip unless: the task is web\u002Fbrowser automation that can run in the\nworkspace. Do not attach to host Chrome, Playwright, or Computer Use as a\nshortcut.\n\n## Do NOT\n\n- Do not dump all MCP tools into the agent context at startup.\n- Do not send users to generic MCP\u002Fconfiguration pages for Codex for Linux\n  Agent Workspace setup.\n- Do not start a workspace without `ack_hidden_workspace=true` and a `purpose`.\n- Do not send input to or screenshot the user's real desktop.\n- Do not perform purchases, checkout, account changes, or sends without explicit\n  user approval.\n- Do not leave workspaces running after the task.\n\n## Example Trajectory\n\nTask: \"QA the settings dialog of myapp.\"\n\n1. `mcp_agent_context` and `mcp_task_plan` for app-QA.\n2. `mcp_permissions` to confirm the active boundary.\n3. `workspace_start` with `ack_hidden_workspace=true` and purpose\n   `\"QA myapp settings\"`.\n4. `workspace_launch_app` with the app command.\n5. `workspace_observe` to find the window.\n6. `workspace_click` \u002F `workspace_type_text` to exercise Settings.\n7. `workspace_screenshot` and `workspace_read_app_log` for evidence.\n8. `workspace_stop`.\n\n## Constraints\n\n- Keep activation low-noise: only use this skill for background GUI\u002Fbrowser work.\n- Keep context small: frontmatter loads by default; body loads on activation;\n  tool schemas load on demand.\n- Omit `allowed-tools` intentionally: different hosts namespace MCP tools\n  differently, and the skill routes the full family progressively.\n- Validate with `agnix --target generic|claude-code|cursor|codex|kiro`\n  when changing this file.\n",{"name":5,"description":7},"main","eadfc3835660b00f1d9e2ed35cd8e1c6867da1ab","linux-features\u002Fagent-workspace\u002Fskills\u002Fagent-workspace-linux",4,"https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux","https:\u002F\u002Fgithub.com\u002Fdistsystem","2026-10-30 14:37:17",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add distsystem\u002Fcodex-desktop-linux --skill agent-workspace-linux",{"tool":36,"command":37},"bunx","bunx skills add distsystem\u002Fcodex-desktop-linux --skill agent-workspace-linux",{"tool":39,"command":40},"pnpm","pnpm dlx skills add distsystem\u002Fcodex-desktop-linux --skill agent-workspace-linux","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux\u002Fblob\u002Fmain\u002Flinux-features\u002Fagent-workspace\u002Fskills\u002Fagent-workspace-linux\u002FSKILL.md",true,[],[45,52,59,66,73,80],{"skillNo":46,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":18},"01M74FR9A410W07GW6YSW1V26H","Unsafe Symlink","Use when testing symlink safety.","https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Funsafe-symlink\u002FSKILL.md",[],[],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":18},"01M74FR7164RYY6XW24HJTHHMB","Mac Only","Use when automating Finder with AppleScript.","https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Fplatform-macos\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":18},"01M74FR51BWCJA98GEKME57H01","Instruction Only","Use when answering with reusable written instructions.","https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Finstruction-only\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":18},"01M74FR31K404DC61C2EREYZ2X","Desktop Act","Use when clicking and typing in a Linux desktop app.","https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Fdesktop-act\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":18},"01M74FR1MABEK5FT6EBQDD6XF9","Collision Tripwire","Use when testing script handling.","https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux\u002Fblob\u002Fmain\u002Frecord-replay-linux\u002Ftests\u002Ffixtures\u002Fskills\u002Fcollision-tripwire\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":18},"01M74FQZF9RK9ETY739V4C4RMK","record-and-replay","Use when the user wants Codex to record a Linux desktop or browser workflow and turn it into a reusable skill. Requires the Record & Replay event-stream MCP server.","https:\u002F\u002Fgithub.com\u002Fdistsystem\u002Fcodex-desktop-linux\u002Fblob\u002Fmain\u002Flinux-features\u002Frecord-and-replay\u002Fplugin-template\u002Fskills\u002Frecord-and-replay\u002FSKILL.md",[],[]]