The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill utilizes a project discovery agent (uipath-project-discovery-agent.md) that reads authored source code (.cs) and workflow files (.xaml) to generate context. This creates a surface for indirect prompt injection if an attacker embeds malicious instructions in the code comments or metadata of a project being analyzed.
Ingestion points: The agent reads project files using Glob, Grep, and Read tools (e.g., **/*.xaml, **/*.cs).
Boundary markers: The skill instructions recommend wrapping external content in delimiters with warnings, though the discovery agent lacks explicit sanitization logic for read content.
Capability inventory: The skill can execute shell commands via the uip CLI and perform file system operations (Write, Edit).
Sanitization: No explicit sanitization or filtering of file content was observed during the discovery phase.
[DYNAMIC_EXECUTION]: The skill provides documentation and instructions for activities that execute external or dynamic code, such as Execute Script (JavaScript), Invoke VBA, Execute Macro, and Invoke Code. These are standard features of the UiPath platform for extensibility.
[COMMAND_EXECUTION]: The skill frequently executes the uip CLI tool to manage projects (e.g., uip rpa init, uip rpa build, uip rpa run). These operations are intended for automation development and management.
[EXTERNAL_DOWNLOADS]: The skill manages project dependencies and entity definitions through official registries (NuGet/Data Service). It includes specific rules to prevent silent installation of untrusted marketplace templates.

uipath-rpa