Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

peekaboo-cli

v1.0.0

macOS UI automation CLI tool for screen capture, window control, element clicking, text input, and more. Use when users need macOS desktop automation, UI tes...

0· 113·0 current·0 all-time
byNEE@terryso
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Benign
medium confidence
Purpose & Capability
Name/description (macOS UI automation) align with the provided SKILL.md and many reference docs. The skill is instruction-only and relies on the Peekaboo CLI; it does not request unrelated credentials or binaries. The allowed tool set (Bash(peekaboo:*)) matches the declared purpose.
Instruction Scope
SKILL.md and the references consistently instruct running peekaboo CLI commands (see, click, type, capture, clipboard, config, agent, etc.). These are within the stated scope (desktop automation and UI capture). However many commands allow saving or emitting screenshots, reading/restoring the clipboard, and reading/modifying Peekaboo config (~/.peekaboo/), which are sensitive operations; the instructions do not limit or sanitize what gets captured or transmitted.
Install Mechanism
This is an instruction-only skill with no install spec or embedded code; that is the lowest install risk. The README points users to install Peekaboo via Homebrew/GitHub (expected for a CLI), which is usual for this type of tool.
!
Credentials
The skill declares no required environment variables, which is appropriate. But the Peekaboo CLI exposed by the skill can access highly sensitive local state: screen captures, the system clipboard, saved snapshots, and the Peekaboo config (which may contain API keys or provider credentials). Although those accesses are coherent with UI automation, they materially increase risk if the skill or agent is untrusted because they enable exfiltration of private data.
Persistence & Privilege
Skill flags show no special persistence (always:false) and model invocation is allowed (default). The skill does not request to modify other skills or system-wide agent settings. Autonomous invocation is platform-default — not a new privilege introduced here.
Assessment
This skill is internally consistent for macOS UI automation: it expects you to have the Peekaboo CLI and macOS accessibility/screen-recording permissions. Before installing or enabling it, confirm you trust the skill's publisher (no homepage/source is provided here). Key things to consider: - Sensitive access: Peekaboo can capture full-screen screenshots, read/write clipboard contents, drive keystrokes, and read ~/.peekaboo config (which may contain API keys). Treat those capabilities as high-risk — they can expose secrets or private content. - Source integrity: The skill metadata shows no homepage/source. Prefer installing Peekaboo from the official repo or Homebrew tap linked in the docs (verify the tap and repository). Inspect the skill files you copy into ~/.openclaw/ or ~/.claude/ before enabling. - Least privilege: Only grant macOS Screen Recording and Accessibility permissions if you trust the tool. Consider granting permissions to a dedicated, sandboxed terminal or temporary account if you need to limit exposure. - Operation mode: Use dry-run or interactive modes first (e.g., peekaboo agent --dry-run or --chat) to confirm what the skill will run before allowing autonomous runs. Avoid enabling broad autonomous invocations for skills that can capture screens/clipboard unless publisher and code are verified. - Audit outputs: Be cautious of commands that send captured images or metadata to external AI providers (the CLI supports an "--analyze" flag and provider config). Review ~/.peekaboo/config and provider entries to ensure no accidental outward transmission of sensitive screenshots or credentials. If you need stronger assurance, obtain the upstream source (GitHub repo), inspect the CLI/tooling, and confirm the skill files haven't been tampered with. If you cannot verify the origin, treat the skill as untrusted despite internal coherence.

Like a lobster shell, security has layers — review code before you run it.

latestvk970232hmarpde9b96sy2h3pr183hgmg

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments