Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
xiaoya-auto-doing
v1.0.0Automate login and page capture for WHUT AI Augmented sites using agent-browser. Use when handling whut.ai-augmented.com or its subdomains, especially to ope...
⭐ 0· 113·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The Python script's behavior matches the stated purpose (open pages, dismiss popups, fill credentials, capture page text). However the skill metadata declares no required binaries while the script calls an external CLI 'agent-browser' — that required binary is not listed. The SKILL.md also references a convenience wrapper 'scripts/whut-open' that is not present in the file manifest. These packaging/declaration mismatches are inconsistent with the stated purpose.
Instruction Scope
Runtime instructions and the script limit their actions to driving agent-browser, reading credentials from environment or a local secret file, and writing a JSON page dump. The script does not send captured data to remote endpoints itself. This scope is consistent with automating browser login and capture, but it will collect full page text (potentially including sensitive info) and save it to latest_page_dump.json — users should expect capture/exfiltration of page content to local disk.
Install Mechanism
There is no install spec (instruction-only plus one script file). Nothing is downloaded or extracted during install. This is lower risk, but the script requires the agent-browser CLI at runtime which the package does not declare.
Credentials
The script legitimately needs WHUT_USERNAME/WHUT_PASSWORD or a path to a local secret file. Those credential sources are described in SKILL.md and used by the code, but the skill metadata does not declare any required environment variables. Also the skill suggests storing secrets in a file under the skill folder (./local/whut_ai_secret.json) which is a poor practice — packaging or placing credentials inside a skill folder increases risk of accidental disclosure.
Persistence & Privilege
The skill is not always-enabled and does not request elevated platform privileges. It writes runtime output and may create/read a local secret file within its own folder, but it does not modify other skills or global configuration.
What to consider before installing
This skill appears to implement browser-driven login and page capture for whut.ai-augmented.com, but there are a few red flags to consider before installing or using it:
- The script calls the external 'agent-browser' CLI, but the skill metadata does not list that binary as required. Verify you have a trusted agent-browser binary installed and understand its permissions.
- SKILL.md mentions a 'scripts/whut-open' wrapper that is not present in the provided file manifest. Confirm which file you should run (scripts/auto_login.py is present and appears to be the main script).
- The skill reads credentials from WHUT_USERNAME/WHUT_PASSWORD or from a local JSON file. Do NOT store real credentials inside the skill package. Prefer environment variables or a secure secret store, and use a throwaway/test account when first trying the skill.
- The tool captures full page text and saves it to latest_page_dump.json, which may include sensitive or personally identifiable information. Treat those outputs as sensitive local data.
- Because the code uses subprocess to call agent-browser, inspect and trust that CLI (and its network/browser behavior) before running. Run the skill in a restricted environment (e.g., dedicated VM or container) if you are uncertain about origin/trust.
If you want to proceed safely: (1) confirm/put agent-browser on PATH from a trusted source, (2) do not place real credentials in the skill folder, (3) test with a non-sensitive account first, and (4) inspect runtime outputs for unexpected network calls or sensitive data being written or transmitted.Like a lobster shell, security has layers — review code before you run it.
latestvk97c966xcxthasbbcp9yvt591983bsvd
License
MIT-0
Free to use, modify, and redistribute. No attribution required.