The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill integrates with an MCP server to inspect and interact with the UI tree of a running application, which presents a surface for indirect prompt injection where a malicious application could provide instructions designed to influence agent behavior. \n
Ingestion points: Accessing element trees and properties through the get_element_tree and get_element_properties tools as documented in reference/debugging-and-mcp.md. \n
Boundary markers: There are no explicit instructions to treat data from the MCP server as untrusted or to use delimiters to separate it from agent instructions. \n
Capability inventory: The agent is authorized to execute shell commands (slint-viewer, cargo, npm, uv) and perform local network operations (curl to the MCP endpoint) as described in SKILL.md and reference/debugging-and-mcp.md. \n
Sanitization: The skill does not mention any validation or sanitization protocols for the UI data retrieved from external processes. \n- [EXTERNAL_DOWNLOADS]: The skill documents procedures for downloading and installing official development tools and project templates from the vendor's GitHub organization. \n
Fetches prebuilt binaries for slint-viewer and slint-lsp from the official slint-ui repository releases. \n
References standard project templates for Rust, C++, Node.js, and Python maintained by the vendor on GitHub. \n- [PRIVILEGE_ESCALATION]: The tools-install.md file includes standard instructions for installing system dependencies on Linux using sudo apt install. \n- [COMMAND_EXECUTION]: The skill guides the agent in using CLI tools for UI verification (slint-viewer --check) and managing project dependencies through cargo, npm, and uv.

slint