File contents Agent Linux Control
One CLI for Linux OS control. Prefer DOM/app APIs for web content; use this for desktop UI, browser chrome, dialogs, downloads, file pickers, screenshots, visual verification.
Loop
agent-linux-control doctor
agent-linux-control manifest --brief
agent-linux-control observe --brief --output /tmp/alc.png
Inspect screenshot.
Act once, or batch with sequence --brief.
Verify via wait-change --brief or observe --brief.
Recover/debug via journal tail.
Token Rules
Prefer --brief for agent loops. Use full JSON only when diagnosing.
Prefer paste over type for long text.
Prefer sequence --brief for known multi-step plans; fewer tool calls, fewer tokens.
If alc-daemon is running, CLI and MCP input commands auto-use it; otherwise Python /dev/uinput fallback runs.
Prefer MCP manifest/observe/sequence with {"brief":true} when MCP available.
Journal redacts text payloads to length + SHA-256.
Prefer app-native APIs for real work inside apps; use desktop control for launch, dialogs, visual verification.
For long autonomous work, prefer isolated workspace when available; direct desktop shares user pointer/focus.
Commands
Contract: agent-linux-control manifest --brief
Observe: agent-linux-control observe --brief --output /tmp/alc.png
Watch: agent-linux-control watch --brief --count 10 --interval 0.5 --output-dir /tmp/alc-watch
Wait: agent-linux-control wait-change --brief --timeout 10 --output /tmp/alc-changed.png
Screenshot: agent-linux-control screenshot --output /tmp/alc.png
Move: agent-linux-control move 200 0
Goto: agent-linux-control goto 900 600
Click: agent-linux-control click left --x 900 --y 600
Scroll: agent-linux-control scroll -5
Type: agent-linux-control type "hello"
Paste: agent-linux-control paste "hello"
Key: agent-linux-control key enter
Hotkey: agent-linux-control hotkey ctrl+l
Clipboard: agent-linux-control clipboard set "text" / agent-linux-control clipboard get
Browser: agent-linux-control browser --prefer chrome --require-prefer https://example.com
Batch: agent-linux-control sequence --brief --file plan.json
Daemon bench: agent-linux-control bench daemon --count 8
MCP: agent-linux-control mcp
Journal: agent-linux-control journal tail --lines 10
Obsidian: agent-linux-control brain export --output-dir ./agent-linux-control-vault
Daemon: alc-daemon serve --socket /tmp/agent-linux-control.sock
Safety
Never type secrets, approve elevated prompts, submit payments, send messages, or delete data unless user explicitly asked for that exact action.
Observe before/after important UI actions.
If click misses, re-observe and recalc; do not repeat blindly.
Wayland goto is approximate.
If /dev/uinput not writable, ask user to run installer or enable udev rule. Do not bypass permissions.
Install
curl -fsSL https://raw.githubusercontent.com/antharmaya/agent-linux-control/main/install.sh | sh
Local:
AGENT_LINUX_CONTROL_SOURCE_DIR="$PWD" ./install.sh
Source: antharmaya/agent-linux-control — distributed by TomeVault .
1 --- 2 name: antharmaya-agent-linux-control-agent-linux-control 3 description: Agent Linux Control 4 --- 5 6 # Agent Linux Control 7 8 One CLI for Linux OS control. Prefer DOM/app APIs for web content; use this for desktop UI, browser chrome, dialogs, downloads, file pickers, screenshots, visual verification. 9 10 ## Loop 11 12 1. `agent-linux-control doctor` 13 2. `agent-linux-control manifest --brief` 14 3. `agent-linux-control observe --brief --output /tmp/alc.png` 15 4. Inspect screenshot. 16 5. Act once, or batch with `sequence --brief`. 17 6. Verify via `wait-change --brief` or `observe --brief`. 18 7. Recover/debug via `journal tail`. 19 20 ## Token Rules 21 22 - Prefer `--brief` for agent loops. Use full JSON only when diagnosing. 23 - Prefer `paste` over `type` for long text. 24 - Prefer `sequence --brief` for known multi-step plans; fewer tool calls, fewer tokens. 25 - If `alc-daemon` is running, CLI and MCP input commands auto-use it; otherwise Python `/dev/uinput` fallback runs. 26 - Prefer MCP `manifest/observe/sequence` with `{"brief":true}` when MCP available. 27 - Journal redacts text payloads to length + SHA-256. 28 - Prefer app-native APIs for real work inside apps; use desktop control for launch, dialogs, visual verification. 29 - For long autonomous work, prefer isolated workspace when available; direct desktop shares user pointer/focus. 30 31 ## Commands 32 33 - Contract: `agent-linux-control manifest --brief` 34 - Observe: `agent-linux-control observe --brief --output /tmp/alc.png` 35 - Watch: `agent-linux-control watch --brief --count 10 --interval 0.5 --output-dir /tmp/alc-watch` 36 - Wait: `agent-linux-control wait-change --brief --timeout 10 --output /tmp/alc-changed.png` 37 - Screenshot: `agent-linux-control screenshot --output /tmp/alc.png` 38 - Move: `agent-linux-control move 200 0` 39 - Goto: `agent-linux-control goto 900 600` 40 - Click: `agent-linux-control click left --x 900 --y 600` 41 - Scroll: `agent-linux-control scroll -5` 42 - Type: `agent-linux-control type "hello"` 43 - Paste: `agent-linux-control paste "hello"` 44 - Key: `agent-linux-control key enter` 45 - Hotkey: `agent-linux-control hotkey ctrl+l` 46 - Clipboard: `agent-linux-control clipboard set "text"` / `agent-linux-control clipboard get` 47 - Browser: `agent-linux-control browser --prefer chrome --require-prefer https://example.com` 48 - Batch: `agent-linux-control sequence --brief --file plan.json` 49 - Daemon bench: `agent-linux-control bench daemon --count 8` 50 - MCP: `agent-linux-control mcp` 51 - Journal: `agent-linux-control journal tail --lines 10` 52 - Obsidian: `agent-linux-control brain export --output-dir ./agent-linux-control-vault` 53 - Daemon: `alc-daemon serve --socket /tmp/agent-linux-control.sock` 54 55 ## Safety 56 57 - Never type secrets, approve elevated prompts, submit payments, send messages, or delete data unless user explicitly asked for that exact action. 58 - Observe before/after important UI actions. 59 - If click misses, re-observe and recalc; do not repeat blindly. 60 - Wayland `goto` is approximate. 61 - If `/dev/uinput` not writable, ask user to run installer or enable udev rule. Do not bypass permissions. 62 63 ## Install 64 65 ```sh 66 curl -fsSL https://raw.githubusercontent.com/antharmaya/agent-linux-control/main/install.sh | sh 67 ``` 68 69 Local: 70 71 ```sh 72 AGENT_LINUX_CONTROL_SOURCE_DIR="$PWD" ./install.sh 73 ``` 74 75 --- 76 > Source: [antharmaya/agent-linux-control](https://github.com/antharmaya/agent-linux-control) — distributed by [TomeVault](https://tomevault.io). 77 <!-- tomevault:4.0:skill_md:2026-06-16 -->
tomevault-io/skills-registry/tree/main/antharmaya--agent-linux-control--agent-linux-control commit 1227648d42
Frequently asked questions How do I install the Antharmaya Agent Linux Control Agent Linux Control skill? Run npx get-skill add tomevault-io/antharmaya-agent-linux-control-agent-linux-control in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
What does the Antharmaya Agent Linux Control Agent Linux Control skill do? Agent Linux Control It is listed under AI & ML on SkillMD.
Is Antharmaya Agent Linux Control Agent Linux Control safe to use? SkillMD's automated safety review verdict for this skill is WARNING. Independent scanners report: SkillSpector: PASS, Skill Scanner: WARNING. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
Which AI agents work with Antharmaya Agent Linux Control Agent Linux Control? This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Is Antharmaya Agent Linux Control Agent Linux Control free to use? Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
Who published Antharmaya Agent Linux Control Agent Linux Control? tomevault-io (@tomevault-io) published this skill. Their other Agent Skills are listed on their SkillMD profile.