desktop-automation-100per100-local

WarnAudited by ClawScan on May 18, 2026.

Overview

This appears to be a real local desktop automation skill, but it gives an agent broad control of your screen, keyboard, mouse, clipboard, and macros, and some typed content may be stored locally in logs.

Install this only if you truly need full local desktop automation. Keep dry-run enabled for testing, require explicit approval before real clicks or typing, avoid recording or typing secrets, review/delete logs and macro files, and run it in a constrained profile or VM when possible.

Findings (5)

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 could click buttons, type into apps, submit forms, change settings, or replay a macro in whatever desktop context is active.

Why it was flagged

The skill intentionally exposes real mouse/keyboard and macro execution over the whole desktop. This fits the stated purpose, but it is broad, high-impact authority and the artifacts do not show a mandatory user-approval gate before non-dry-run actions.

Skill content
Automate desktop interactions without APIs: ... Click, type, drag, scroll ... Record and replay macros
Recommendation

Use only when you are present and can approve real actions; prefer dry_run first, keep sensitive apps closed, and consider running it in a separate OS profile or VM.

What this means

Sensitive text entered by the automation may remain in local log files even if you did not intentionally record a macro.

Why it was flagged

The type action logs the exact text it sends. Since the skill also documents persistent audit logging, passwords, tokens, or other sensitive text typed through the skill could be retained locally in logs.

Skill content
logger.info("Typed: %s", text)
Recommendation

Do not use the type action for passwords, tokens, or private messages unless logging is disabled or redacted; periodically review and delete automation logs.

What this means

Recorded macros may contain private text, window activity, or credentials if recording is left on during sensitive work.

Why it was flagged

Macro recording is clearly disclosed and purpose-aligned, but it can store raw keystrokes and mouse activity, including accidental secrets, in persistent local files.

Skill content
This skill captures ALL keyboard and mouse events. ... Recorded macros are stored as JSON in `recorded_macro/` directory
Recommendation

Only record non-sensitive workflows, store macro files securely, and delete recordings that may contain private input.

What this means

The skill runs local code with the permissions of the current user account.

Why it was flagged

The Node entrypoint launches a local Python script. This is expected for the skill's Python automation implementation, but users should understand that installing the skill allows local Python code execution.

Skill content
const result = spawnSync('python', args, { encoding: 'utf-8' });
Recommendation

Install only from a source you trust and review the local files before enabling the skill.

What this means

Future package versions could change behavior or introduce dependency risk.

Why it was flagged

The requirements file uses lower-bound version ranges rather than exact pins, while the documentation elsewhere discusses tested versions. This can lead to dependency drift, though the packages are purpose-aligned.

Skill content
pyautogui>=0.9.53
pygetwindow>=0.0.9
Pillow>=8.0.0
opencv-python>=4.5.0
Recommendation

Use a virtual environment and consider pinning exact dependency versions before installation.