The Agent Skills Directory

[COMMAND_EXECUTION]: The skill initializes a local Model Context Protocol (MCP) server by executing the command stored in the DOCUMENT_MCP_CMD environment variable. This is a standard operational requirement for the skill's functionality and follows platform conventions for tool integration.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from the NetClaw system, which presents a potential injection surface. To mitigate this risk, it enforces a strict JSON schema (requiring 'v' and 'src' fields) and includes a specific sanitization rule to prevent spreadsheet formula injection by ensuring descriptions beginning with '=' are treated as literal text.
[DATA_EXPOSURE]: The skill is designed to handle network and ticket data for reporting purposes. It implements strong security controls by mandating per-element provenance and source tagging, which prevents data fabrication and ensures all generated content is traceable to an official source.

document-generation