property-utility-apportionment-check-free
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill includes instructions in
SKILL.mdfor the agent to execute a command that downloads external code. - Evidence: The provided prompt template instructs the AI to run
npx -y @alipay/agent-payment@latest install. This command fetches and executes the@alipay/agent-paymentpackage from the npm registry, which is a well-known service. - [PROMPT_INJECTION]: The
SKILL.mdfile contains a pre-formatted text block intended to be processed by the AI agent to initiate an upsell flow. - Evidence: The text instructs the agent to install specific software and proceed with a purchase, using steering language such as "我已授权安装依赖和发起购买" (I have authorized the installation of dependencies and initiation of purchase) to guide the agent's actions.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external text data (property utility bills) provided by the user.
- Ingestion points: The
textparameter in therunfunction withinscripts/engine/property-utility-apportionment-check.js. - Boundary markers: The data is parsed using tab or multi-space delimiters to identify columns in a table structure.
- Capability inventory: The skill is configured with access to the
BashandReadtools via its YAML frontmatter. - Sanitization: The engine performs arithmetic validation and numerical parsing on the input data, treating the content as data for calculations rather than executable instructions, which reduces the risk of the input influencing the script's execution logic.
Audit Metadata