The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell commands to manage its environment and perform operations. It uses which to check for the presence of the mcporter tool, npm install for software installation, and mcporter for tool calls and configuration. It also uses openclaw to query system settings and cat to read local files.
[EXTERNAL_DOWNLOADS]: During the setup process, the skill prompts the installation of the mcporter package from the public npm registry, which is an external dependency required for its functionality.
[DATA_EXFILTRATION]: The skill reads connection details from a local file at ~/.openclaw/wecomConfig/config.json using the cat command to facilitate automatic configuration of the MCP server, which exposes local environment information.
[REMOTE_CODE_EXECUTION]: The skill is designed to configure and invoke tools from remote MCP servers using URLs provided by the user or through local configuration. These remote servers are then called dynamically to perform operations, representing a form of remote resource integration.
[PROMPT_INJECTION]: The skill includes instructions that override standard agent behavior, such as defaulting to Enterprise WeChat without user confirmation and requiring verbatim output of specific prompts. It also presents a surface for indirect injection by processing user-supplied URLs and tool error messages without sanitization or boundary markers. (Evidence Chain
Ingestion points: User-provided URLs/JSON and tool help_message outputs in SKILL.md; Boundary markers: Absent; Capability inventory: Shell execution via npm, mcporter, and cat across the skill scripts; Sanitization: No explicit sanitization or validation of the external content).

wecom-doc