The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents components and APIs that create an attack surface for indirect prompt injection by processing potentially untrusted data alongside sensitive agent capabilities.
Ingestion points: Data enters via the Input, TextArea, and OTP components in references/zaui-form.md, as well as device APIs for scanning QR codes (scanQRCode) and NFC tags (scanNFC) in references/api-device.md.
Boundary markers: The documentation does not provide explicit delimiters or instructions to treat data from these ingestion points as untrusted.
Capability inventory: The skill enables network operations (openWebview in references/api-ui.md, openChat in references/api-zalo.md), device control (getLocation, takePhoto in references/api-device.md), and data persistence (setItem in references/api-storage.md).
Sanitization: No explicit sanitization or validation routines for external content are detailed in the provided references.
[DYNAMIC_EXECUTION]: The skill provides code examples in references/react-best-practices.md that use dangerouslySetInnerHTML to inject and execute a JavaScript block in the DOM for theme initialization.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to download and install command-line tools and libraries from the Zalo platform ecosystem.
Evidence: npm install -g zmp-cli and npm install zmp-ui zmp-sdk as documented in SKILL.md.
Note: These resources originate from Zalo, a well-known technology service provider.
[COMMAND_EXECUTION]: The skill provides usage instructions for the zmp command-line utility to initialize and manage projects.
Evidence: zmp create my-app && cd my-app && zmp start in SKILL.md and references/getting-started.md.

zalo-mini-app