T08 · Insecure Dependencies
- Location
SKILL.md:50- Finding
Unpinned Remote Repository and Unverified Third-Party Dependencies
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This skill is for real Booking.com account automation, but it asks users to install unpinned external code that handles credentials, persistent sessions, guest messages, and live rate changes.
Review this before installing. Use only a pinned, audited commit of the external repository, lock and verify dependencies, protect the .env and .chrome-data files, and require explicit human approval with a preview before sending guest messages or changing rates.
SKILL.md:50Unpinned Remote Repository and Unverified Third-Party Dependencies
The skill is designed to access and use Booking.com credentials from a local .env file and persist an authenticated browser session, which materially increases credential exposure and session theft risk on the host. Although this is part of intended functionality rather than overtly malicious behavior, it still constitutes sensitive credential handling with elevated impact because it grants access to a live partner extranet account and potentially 2FA-backed sessions.
- browser_automation: Uses Google Chrome via remote debugging (CDP on localhost:9222)
- local_storage: Persists browser session in .chrome-data/ directory to avoid repeated login/2FA
- network: Connects only to admin.booking.com and account.booking.com
- filesystem: Reads .env for credentials, writes Excel files to ./downloads/
---
# Booking.com Extranet Manager
The skill enables sending messages to Booking.com guests but does not explicitly warn that this action affects real external users and may be irreversible once sent. In an agent setting, omission of that warning increases the risk of accidental customer communication, reputational harm, and unauthorized or mistaken responses.
The rate update command is described as a routine operation without clearly stating that it changes live pricing data for an actual property. In an autonomous or semi-autonomous environment, missing this warning can lead to unintended price changes, revenue loss, booking disruption, or contractual/customer issues.
No suspicious patterns detected.