desktop mouse
PassAudited by VirusTotal on May 12, 2026.
Overview
Type: OpenClaw Skill Name: molt-mouse Version: 1.0.0 The skill provides direct control over the user's mouse via the `ydotool` utility, as implemented in `molt-mouse.txt`. While the `SKILL.md` instructions for the AI agent are well-constrained to only execute `molt-mouse` commands and the script itself appears robust against shell injection, the inherent capability to control a user's desktop input is a high-risk behavior. This capability, even when implemented without malicious intent, could lead to unintended actions if the agent's instructions or reasoning were compromised, making it suspicious.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If invoked with the wrong coordinates or at the wrong time, the agent could click or move over unintended UI elements.
The skill intentionally gives the agent a host-exec path to issue local mouse commands. This is purpose-aligned and bounded to the molt-mouse wrapper, but mouse clicks can affect the active desktop.
Use the exec tool with host=gateway. ONLY run commands that start with: `molt-mouse ...`
Install only if you want the agent to control the local pointer, and prefer explicit coordinates/actions when asking it to click.
A user reviewing only SKILL.md may not notice that the helper can also perform holds and drags.
The wrapper supports hold and drag operations beyond the shorter Supported list in SKILL.md. These are still mouse-control functions, but the executable's practical scope is slightly broader than the runtime summary.
cmds: move <dx> <dy> abs <x> <y> click [left|right|middle] hold [left|right|middle] <seconds> drag <dx> <dy> [left|right|middle] [hold_seconds] drag-smooth <dx> <dy> [steps] [left|right|middle]
Document all supported mouse actions in SKILL.md or ensure the agent only uses the listed commands.
If a different local `molt-mouse` binary is present, the agent would run that binary rather than a reviewed installed artifact.
The skill expects a local `molt-mouse` command but does not provide an installation path in the install spec, so users need to confirm what binary will actually run.
No install spec — this is an instruction-only skill.
Verify the local `molt-mouse` command path and contents before enabling the skill.