atlas-flight-booking

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill automatically installs the uv package manager from the official Astral domain (astral.sh) if it is not present on the host system. This is used exclusively for environment setup and tool management.\n- [REMOTE_CODE_EXECUTION]: The skill uses piped shell commands (curl | sh, wget | sh) and PowerShell execution (Invoke-Expression) to install uv. While these are sensitive patterns, they are targeting a well-known, legitimate technology domain and are used for standard software installation as part of the skill's primary function.\n- [COMMAND_EXECUTION]: The skill manages the installation and execution of the atlas-flight-booking CLI (version 0.3.12). The agent is instructed to interact with this tool using a strict JSON-based contract, ensuring predictable behavior and structured data handling.\n- [PRIVILEGE_ESCALATION]: On Windows systems, the skill utilizes the ExecutionPolicy ByPass flag to execute the official uv installer. This is a standard and expected procedure for bootstrapping the required CLI environment.\n- [DATA_EXPOSURE]: The skill handles sensitive passenger information, such as document numbers and contact details. It implements significant security controls to protect this data: it directs the agent to provide data via standard input (stdin) rather than command-line arguments to prevent exposure in process lists or shell history; it explicitly prohibits logging, saving, or echoing personal data back to the user; and it requires explicit human-in-the-loop approval for payment summaries and price increases.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes data from flight search results via the CLI. To mitigate potential injection risks from tool outputs, the instructions mandate that the agent branch only on structured response codes rather than natural language messages and present only normalized fields to the user.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 08:56 AM
Security Audit — agent-trust-hub — atlas-flight-booking