browserskill-release-test

Pass

Audited by Gen Agent Trust Hub on Nov 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The release_audit.py script executes a local BrowserSkill binary to perform automated browser testing. The commands and arguments are derived from test plans, and the process is managed using Python's subprocess module without shell expansion.
  • [EXTERNAL_DOWNLOADS]: The skill benchmark tasks involve visiting and extracting data from well-known domains like GitHub, Bilibili, and Taobao. These operations are intended for auditing the tool's performance on real-world websites.
  • [INDIRECT_PROMPT_INJECTION]: The skill explicitly tests for indirect prompt injection resilience. It includes HTML fixtures with hidden text, transparent layers, and subpixel font-size blocks (steganography) to verify that the auditing process can effectively identify and strip non-visible instructions.
  • [DYNAMIC_EXECUTION]: The skill uses a local server (serve.py) to host test fixtures on the loopback address and dynamically generates test results and benchmarking metrics based on runtime execution data.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 13, 2026, 05:38 PM