volcengine-sale

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill includes logic to automatically provision required tools if they are not already available in the environment.
  • Evidence: SKILL.md describes an automated installation process ("依赖 skill 自动加载") that triggers when a requested product's skill is missing.
  • Source URLs: The skill references official vendor repositories such as https://github.com/volcengine/ark-cli and https://github.com/volcengine-tls/ve-tls-cli in references/product-rule.md as sources for git clone operations.
  • [COMMAND_EXECUTION]: The skill performs shell-based operations to execute billing actions and check instance status via the vendor's CLI tool.
  • Evidence: references/commonbuy.md outlines the execution of ve billing CommonBuy and ve billing ListAvailableInstances using the Bash tool.
  • Security Controls: The instructions mandate that the agent must present the command to the user and obtain explicit permission before execution ("未经用户确认不得执行").
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as a router for user purchasing intents, which involves interpreting user text to select internal routing rules and configuration parameters.
  • Ingestion points: User requests involving product names and action verbs (e.g., "开通", "购买") defined in SKILL.md and references/product-rule.md.
  • Boundary markers: A mandatory user confirmation step serves as a boundary between the interpretation of the request and the execution of tool commands.
  • Capability inventory: The skill has the ability to execute Bash commands including ve, git, and uuidgen.
  • Sanitization: Command parameters are primarily derived from static local configuration files (e.g., references/tos.md) rather than being directly interpolated from untrusted user strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 06:46 AM
Security Audit — agent-trust-hub — volcengine-sale