The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to extract the agentApiToken from a local configuration file at %LOCALAPPDATA%\OpenQuickHost\appsettings.local.json to authenticate with the local management API.
[REMOTE_CODE_EXECUTION]: The extension host allows for the dynamic compilation and execution of arbitrary C# and PowerShell code provided via manifest files (inline) or local script files. It also supports remote installation of extensions from an app store.
[COMMAND_EXECUTION]: The documentation provides instructions for establishing persistence and performing high-privilege system operations, such as:
Configuring extensions to launch automatically with the host application using the startup manifest parameter.
Implementing resident background services using Win32 API hooks (RegisterHotKey, SetWindowsHookEx) to persist logic after initial execution.
Using PowerShell for system-wide automation, including registry modification and process management.
[PROMPT_INJECTION]: The skill defines a significant attack surface for indirect prompt injection by ingesting untrusted data from multiple sources without sanitization:
Ingestion points: Data retrieved from the system clipboard (Get-Clipboard), external API responses (hitokoto.cn), and the InputText variable.
Boundary markers: No delimiters or safety instructions are used when processing external data.
Capability inventory: The skill possesses full execution capabilities for C# and PowerShell, local file system access, and network request functionality.
Sanitization: No validation, escaping, or filtering of ingested data is described.
[DATA_EXFILTRATION]: The skill combines the ability to read sensitive local tokens and system data (e.g., clipboard, configuration files) with the capability to perform outbound HTTP requests, providing a clear path for data exfiltration.

yanzi-extension-dev