miniprogram-development
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches vendor-scoped packages such as
@cloudbase/npm-launcherand@cloudbase/cloudbase-mcpfrom the official NPM registry to provide Model Context Protocol (MCP) capabilities for cloud resource management.- [REMOTE_CODE_EXECUTION]: Executes vendor-specific integration tools via thenpxpackage runner to facilitate the discovery and invocation of CloudBase cloud functions, database, and storage services.- [COMMAND_EXECUTION]: Interfaces with the official WeChat Developer Tools CLI (wechatide) to automate simulator tasks, console debugging, and project builds within the user's development environment.- [INDIRECT_PROMPT_INJECTION]: The skill processes local project configuration files and source code to automate development tasks. This ingestion surface is managed through the following evidence chain: - Ingestion points: Reads workspace files such as
project.config.json,app.json, and page-level source code. - Boundary markers: Explicitly mandates following the
Change Safety ProtocolandDeployment Gatebefore modifying or deploying project code. - Capability inventory: Utilizes the
wechatideCLI for IDE automation and thenpxrunner for executing vendor-scoped MCP tools. - Sanitization: Sensitive operations are designed to be performed through the official WeChat Developer Tools interface, which provides a layer of user oversight.- [SAFE]: The skill includes security-positive instructions, such as explicitly warning against fetching remote markdown content into the agent context to prevent potential prompt injection from external documentation.
Audit Metadata