Ghosthand-Skill

Security checks across static analysis, malware telemetry, and agentic risk

Overview

Ghosthand-Skill is a coherent instruction-only guide for a local Android-control API, but users should recognize that it can let an agent view and operate phone UI, clipboard, screenshots, and notifications.

This skill appears benign and purpose-aligned for controlling a local Ghosthand Android runtime. Before installing, make sure you trust the Ghosthand runtime itself, review its Android permissions and policy, and give the agent explicit limits for sensitive actions such as purchases, account changes, message sending, or reading private notifications.

Static analysis

No static analysis findings were reported for this release.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Risk analysis

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.

What this means

An agent using this skill may tap, type, scroll, navigate, or otherwise operate the phone UI.

Why it was flagged

These routes allow an agent to interact with and navigate an Android device. This is central to the skill's stated purpose, but it is still high-impact control that users should intentionally authorize.

Skill content
- semantic or coordinate interaction: `/click`, `/tap`, `/input`, `/type`, `/setText`, `/scroll`, `/swipe`, `/longpress`, `/gesture`
- app and navigation control: `/back`, `/home`, `/recents`
Recommendation

Use it only with a trusted Ghosthand runtime and give the agent clear boundaries; require user confirmation for purchases, account changes, destructive actions, or public posting.

What this means

If Ghosthand has broad Android permissions, the agent may inherit practical ability to observe or control sensitive device surfaces.

Why it was flagged

The skill explicitly relies on delegated Android permissions and Ghosthand policy. That is purpose-aligned, but users should understand that Android authorization can expose privileged device functions.

Skill content
A capability is usable only when both are true:

- Android/system authorization exists
- Ghosthand policy allows the capability
Recommendation

Review Ghosthand's Android permissions and policy settings before use, and grant only the capabilities needed for the task.

What this means

The agent may see private on-screen content, copied secrets, messages, or notifications while operating the device.

Why it was flagged

The local API can expose screenshots, clipboard contents, and notifications through the agent/runtime boundary. This is expected for Android automation, but the data may be sensitive.

Skill content
`GET /screenshot` / `POST /screenshot` — screenshot retrieval when visual truth is needed; ...
`GET /clipboard` / `POST /clipboard` — clipboard read/write
`GET /notify` / `POST /notify` / `DELETE /notify` — notification read/post/cancel
Recommendation

Avoid using the skill while unrelated sensitive content is visible or in the clipboard, and restrict Ghosthand access to trusted local agents.

What this means

This review covers the skill instructions, not the provenance or implementation of the external Ghosthand Android runtime.

Why it was flagged

The skill package does not identify an upstream source or homepage. Because the actual Ghosthand runtime is not included in this instruction-only skill, users should verify that runtime separately.

Skill content
Source: unknown
Homepage: none
Recommendation

Install or run Ghosthand only from a trusted source and review its own permissions, policy, and network exposure.