targeted-proxy-access

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads the xray proxy binary from GitHub's official repository (github.com/XTLS/Xray-core). While GitHub is a well-known service, the binary originates from a third-party project.
  • [REMOTE_CODE_EXECUTION]: The scripts/tunnel.py script executes the downloaded xray binary using subprocess.Popen. The binary is used to establish the proxy tunnel, and the script verifies its integrity using SHA-256 checksums fetched during the download process.
  • [COMMAND_EXECUTION]: The skill provides a launcher script (scripts/launch.py) that uses os.execvpe to execute arbitrary commands with proxy environment variables injected. This allows other tools to be transparently routed through the proxy gateway.
  • [PERSISTENCE]: The skill includes functionality to install a systemd user service (proxy-tunnel.service) on Linux. This persistence mechanism allows the gateway to start automatically on login and enables background monitoring and configuration rotation.
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a fetch command that retrieves external web content into the agent's context. This creates a surface for indirect prompt injection where a malicious website could attempt to influence the agent's behavior.
  • Ingestion points: scripts/tunnel.py via cmd_fetch and urllib.request calls.
  • Boundary markers: Absent; the tool retrieves raw content and passes it to the agent.
  • Capability inventory: The skill can execute arbitrary local commands (via the wrapper), write to configuration files, and perform network operations.
  • Sanitization: Content retrieved from external URLs is processed without sanitization or filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 09:59 PM