The Agent Skills Directory

[COMMAND_EXECUTION]: The lifecycle management script scripts/rterm-backend.mjs executes shell commands and binaries using Node.js spawn and spawnSync modules.\n- [COMMAND_EXECUTION]: The Windows service installation script service/install-windows-service.ps1 bypasses PowerShell execution policies using the -ExecutionPolicy Bypass flag.\n- [EXTERNAL_DOWNLOADS]: The skill initiates external software downloads and installations via npm install and npx. It also fetches update metadata from raw.githubusercontent.com and provides instructions for cloning from GitHub.\n- [REMOTE_CODE_EXECUTION]: The backend facilitates remote command execution on hosts via SSH and WinRM. It includes functionality to deploy the aperf performance monitoring tool to remote Linux environments.\n- [REMOTE_CODE_EXECUTION]: The skill implements a persistence mechanism by registering the backend as a system service on Linux (systemd), macOS (launchd), and Windows (Task Scheduler).\n- [CREDENTIALS_UNSAFE]: The skill stores and manages sensitive connection data and authentication tokens in configuration files like settings.json and access-tokens.json. It utilizes an AES-256-GCM vault for encrypted secret storage.\n- [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing untrusted external data from web searches, page fetches, and user-provided YAML workflows.\n
Ingestion points: Untrusted data enters via the userInput parameter, dagu YAML files, and content retrieved by web intelligence tools.\n
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands in the processed data.\n
Capability inventory: The skill has extensive access to system resources, including shell execution, file system modification, and network access.\n
Sanitization: Safety relies on runtime policy enforcement and secondary model reviews rather than input sanitization.

rterm-backend