computer-use-linux
Use when you need to do computer use tasks on linux, for browser only tasks, use playwriter
Skill by firemonster612
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/firemonster612-computer-use-linux
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/computer-use-linux
What's inside
- Skill name
computer-use-linux- Description
- Use when you need to do computer use tasks on linux, for browser only tasks, use playwriter
- License
- MIT
- Instructions
- 53 lines in SKILL.md · view on GitHub
Agent Skills spec issues
- body: contains an absolute path from the author's machine
Files (1)
- SKILL.md
SKILL.md (first lines)
# computer-use-linux One integration point: the **MCP server** `computer-use-linux` (registered in Claude Code and Codex, launched as `computer-use-linux mcp`). It exposes all observation _and_ input tools: `doctor`, `get_app_state`, `list_apps`, `list_windows`, `focused_window`, `screenshot`, `click`, `drag`, `scroll`, `type_text`, `press_key`, `perform_action`, `set_value`, `activate_window`, `move_window`, `resize_window`, `setup_accessibility`, `setup_window_targeting`. Desktop machines only (fedora, cachy). Headless machines (happy, chubbs) have no desktop to control. ## Procedure 1. Start with `get_app_state` (or `doctor` if things look broken). It reports readiness diagnostics. 2. Before targeted input, call `list_windows`/`focused_window` and verify the intended window by title, app id, pid, or wm class. 3. Prefer semantic targeting from `get_app_state` (element indices or role/name/text/states selectors) over raw coordinates. Use coordinates only when there's no useful accessibility tree. 4. For text, prefer `type_text` with an explicit target selector (`window_id`, `pid`, `app_id`, `wm_class`, `title`, `tty`, ...) over relying on current focus. 5. After mutating actions, re-check state (`get_app_state`, `focused_window`, or app-specific readback). 6. Desktop input is stateful, so never run tool calls against this server concurrently. ## If the MCP tools aren't available The server binary doubles as a CLI for observation only (no input): `computer-use-linux state [APP]`, `screenshot`, `windows`, `apps`, `doctor`. Fix the MCP registration rather than working around it for long. ## Setup / troubleshooting ```bash bun install -g @agent-sh/computer-use-linux # installs the `computer-use-linux` binary computer-use-linux doctor | jq .readiness ``` Ready means: `can_build_accessibility_tree`, `can_query_windows`, `can_send_development_input` all true, `blockers: []`. If not: ```bash computer-use-linux setup # enables AT-SPI, input support systemctl --user enable --now ydotoold # per-user input daemon computer-use-linux setup-window-targeting # GNOME Wayland: shell extension for window queries ``` - On GNOME Wayland, log out/in after `setup-window-targeting` installs the shell extension. - Already-running GTK/Qt/Electron apps need a restart after AT-SPI is first enabled. - GNOME may show a one-time portal prompt on the first screenshot. MCP registration (the server speaks standard stdio JSON-RPC, so no wrapper scripts): ```bash # Claude Code claude mcp add --scope user computer-use-linux ~/.bun/bin/computer-use-linux mcp ``` ```toml # ~/.codex/config.toml [mcp_servers.computer-use-linux] command = "/home/efox/.bun/bin/computer-use-linux" args = ["mcp"] ```
From firemonster612/dot-agents, MIT.
More from firemonster612/dot-agents
- 0.559score
review-loop
firemonster612
Independent Claude and GPT review of a finished change, argued finding by finding with the author.
SkillAFreshness: AClaude Code +9 - 0.555score
full-send
firemonster612
Use when the user says full send or full stop.
SkillAFreshness: AClaude Code +9 - 0.534score
codebase-design
firemonster612
Shared vocabulary for designing deep modules.
SkillAFreshness: AClaude Code +9 - 0.518score
cli-subagents
firemonster612
Use when you need to use a model as a subagent that is not supported in your harness natively via the CLI.
SkillBFreshness: BClaude Code +9 - 0.499score
diagnosing-bugs
firemonster612
Diagnosis loop for hard bugs and performance regressions.
SkillBFreshness: BClaude Code +9 - 0.494score
prototype
firemonster612
Build a throwaway prototype to answer a design question.
SkillBFreshness: BClaude Code +9 - 0.494score
domain-modeling
firemonster612
Build and sharpen a project's domain model.
SkillBFreshness: BClaude Code +9 - 0.494score
grilling
firemonster612
Grill the user relentlessly about a plan, decision, or idea.
SkillBFreshness: BClaude Code +9 - 0.491score
maintain-verification-skill
firemonster612
Periodic pass that keeps a project's verification skill and feature map honest: parallel source readers per feature, one live session driving every feature…
SkillBFreshness: BClaude Code +9 - 0.483score
handoff
firemonster612
Compact the current conversation into a handoff document for another agent to pick up.
SkillBFreshness: BClaude Code +9 - 0.474score
human-writing
firemonster612
Write text that reads as authentically human rather than AI-generated.
SkillCFreshness: CClaude Code +9 - 0.469score
playwriter
firemonster612
Control the user own Chrome browser via Playwriter extension with Playwright code snippets in a stateful local js sandbox via playwriter cli.
SkillCFreshness: CClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Aug 15, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/firemonster612/dot-agents
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from firemonster612
- 0.559score
review-loop
firemonster612
Independent Claude and GPT review of a finished change, argued finding by finding with the author.
SkillAFreshness: AClaude Code +9 - 0.555score
full-send
firemonster612
Use when the user says full send or full stop.
SkillAFreshness: AClaude Code +9 - 0.534score
codebase-design
firemonster612
Shared vocabulary for designing deep modules.
SkillAFreshness: AClaude Code +9 - 0.518score
cli-subagents
firemonster612
Use when you need to use a model as a subagent that is not supported in your harness natively via the CLI.
SkillBFreshness: BClaude Code +9 - 0.499score
diagnosing-bugs
firemonster612
Diagnosis loop for hard bugs and performance regressions.
SkillBFreshness: BClaude Code +9 - 0.494score
prototype
firemonster612
Build a throwaway prototype to answer a design question.
SkillBFreshness: BClaude Code +9
Cite this page
computer-use-linux skill. Agent Mart. https://agentmart.dev/skills/firemonster612-computer-use-linux (last updated 2026-10-24)