The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill instructions direct the agent to visit an external website (https://docs.openclaw.ai/channels/index) to extract and apply installation commands for additional channels. \n
Ingestion points: SKILL.md and references/channel-discovery.md reference the external documentation as the source for installation steps.\n
Boundary markers: Absent; the instructions command the agent to 'Extract exactly from docs' and 'Apply installation/configuration on host.'\n
Capability inventory: The skill utilizes powerful capabilities including global npm installation, plugin installation, and root access via SSH (SKILL.md).\n
Sanitization: No sanitization or validation of the commands fetched from the external source is performed.
[REMOTE_CODE_EXECUTION]: Installation of Remote Plugins. The skill installs extensions and plugins directly from external Git repositories and package registries. \n
Evidence: 'openclaw plugins install https://github.com/DingTalk-Real-AI/dingtalk-openclaw-connector.git' in SKILL.md.\n
Note: This is a core functionality of the managed tool (OpenClaw) and targets a vendor-associated repository.
[EXTERNAL_DOWNLOADS]: Runtime Environment Setup. The skill fetches and executes a setup script for Node.js from a well-known service provider. \n
Evidence: 'curl -fsSL https://deb.nodesource.com/setup_20.x | bash -' in SKILL.md.\n
Source: deb.nodesource.com is a well-known distributor of Node.js binaries; the operation is documented neutrally.
[COMMAND_EXECUTION]: High Privilege Operations. The skill requires root access and performs system-level modifications. \n
Evidence: Usage of 'ssh root@', 'apt-get install -y nodejs', and 'systemctl --user' commands in SKILL.md.

alicloud-platform-openclaw-setup