Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
desktop-automation-100per100-local
v2.0.1Automate desktop tasks locally with mouse, keyboard, window control, OCR, and image recognition using Python on Windows/macOS/Linux.
⭐ 0· 302·1 current·1 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
Name/description, declared requirements, and included code match: this is a local desktop automation/macro recorder + player using PyAutoGUI/OpenCV/pytesseract. Required packages and runtime behaviors align with the stated functionality.
Instruction Scope
SKILL.md and the code explicitly capture ALL keyboard and mouse events, record window titles, save macros to disk, and can replay arbitrary sequences of input. Although the skill provides dry_run and 'safe mode' pattern checks, those checks are pattern-based and superficial (string pattern matching like 'rm ' or 'sudo') and can be bypassed by GUI-driven flows (e.g., launching a terminal and sending keystrokes to execute arbitrary commands). Recorded macros and audit logs may contain sensitive data (passwords, credit cards, window titles). The skill does not require or declare any environment credentials, but its ability to synthesize input lets it interact with networked apps to exfiltrate data (even though the skill itself has no network code).
Install Mechanism
No automatic installer is provided (instruction-only install). The README/SKILL.md instructs the user to place the folder and pip install requirements — this is typical and low-risk compared to remote-download installers. The repository/package URLs referenced are GitHub-like; there are no opaque download URLs or archive extraction steps in the install instructions.
Credentials
The skill requests no environment variables or external credentials and lists only local Python dependencies appropriate to desktop automation, OCR, and image recognition. That is proportionate to the stated purpose.
Persistence & Privilege
The skill writes audit logs to ~/.openclaw/skills/desktop-automation-logs/ and saves recorded macros to a recorded_macro/ directory; it can also create encrypted macro files. It does not declare always:true and does not modify other skills. Persisting logs and macros in user home is expected for this type of tool, but these files can contain sensitive data and should be protected (permissions/encryption).
What to consider before installing
This skill appears to do what it claims (local desktop automation) but is inherently high-risk because it records and replays all keyboard and mouse activity. Before installing:
- Don’t record or store any sensitive input (passwords, credit cards, authentication tokens). Recorded macros and logs are stored locally and can contain raw keystrokes and window titles.
- Treat recorded_macro/ and ~/.openclaw/skills/desktop-automation-logs/ as sensitive data stores; restrict filesystem permissions and consider encrypting backups. The skill supports AES-protected macros, but the presence of that feature means users might be tempted to embed secrets — avoid doing so.
- Do not enable autonomous or unattended execution of macros you did not author and review carefully every macro file before playback. A macro that types key sequences can open terminals or web browsers and perform destructive or exfiltrative actions even though the skill has no network calls itself.
- The 'safe mode' is pattern-based and easily bypassed (GUI-driven actions, obfuscated typing, or multi-step sequences). Do not rely on it as a security boundary.
- Prefer testing with dry_run=true and manual supervision. Consider running the skill in a restricted, sandboxed account or VM if you need to run untrusted macros.
- Audit the full source before use (especially files omitted/truncated in the provided dump). If you plan to allow other users or the agent to invoke the skill autonomously, consider disabling autonomous invocation or adding stronger authorization controls.
Confidence is medium because some files were truncated in the provided file dump; a complete line-by-line review of all source files would increase confidence and could surface any hidden network calls or persistence mechanisms not visible in the excerpts.skill.js:14
Shell command execution detected (child_process).
Patterns worth reviewing
These patterns may indicate risky behavior. Check the VirusTotal and OpenClaw results above for context-aware analysis before installing.Like a lobster shell, security has layers — review code before you run it.
latestvk9753x4wg9t3dve9y9wshhsbx582y20x
License
MIT-0
Free to use, modify, and redistribute. No attribution required.