rollinggo-hotel-booking
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The
scripts/install.pyscript downloads platform-specific binaries from the vendor's official GitHub repository (github.com/RollingGo-AI/oauth-hotel-cli-overseas). - [REMOTE_CODE_EXECUTION]: The skill downloads and executes an external binary (
rgg) to interface with its hotel booking API. It also initiates the installation of a Node.js package (@rollinggo/hotel-global) from the public NPM registry. - [COMMAND_EXECUTION]: The skill uses proxy scripts (
scripts/rgg.jsandscripts/rgg.py) to execute shell commands and launch therggCLI tool with dynamically generated arguments. - [INDIRECT_PROMPT_INJECTION]: The skill retrieves hotel names, descriptions, and facility tags from an external API, creating an attack surface where malicious external content could influence the agent's behavior.
- Ingestion points: Hotel data fields returned by the
search-hotelsandhotel-detailsubcommands. - Boundary markers: Absent; external data is directly interpolated into Markdown templates for user display.
- Capability inventory: Subprocess execution for CLI commands (
rgg.js,rgg.py) and network requests (via the CLI). - Sanitization: No explicit sanitization or validation of the text data received from the API is performed before rendering it to the user.
Audit Metadata