The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides commands for immediate script execution from a remote URL (https://higress.ai/hiclaw/install.sh) using bash process substitution (bash <(...)). It also utilizes PowerShell's Invoke-Expression (iex) to download and run a script from the same domain (install.ps1).
[COMMAND_EXECUTION]: Multiple instances of 'npx -y' are used to run Model Context Protocol (MCP) servers (e.g., @modelcontextprotocol/server-filesystem). This pattern automatically downloads and executes packages from the npm registry at runtime without user confirmation.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to add external Helm repositories (higress.io) and deploy infrastructure components via Helm. It also describes a mechanism for loading skills dynamically from a remote Nacos server (nacos-server:8848).
[PROMPT_INJECTION]: The Manager-Workers architecture relies on Matrix rooms for agent communication, which introduces an indirect prompt injection surface. The ingestion point is the Matrix protocol (Matrix rooms), where untrusted data enters the agent context. There are no boundary markers or delimiters mentioned to isolate instructions. The capability inventory is significant, including subprocess calls via npx and kubectl, and no sanitization or filtering of external room content is described.

hiclaw-multi-agent-collaboration