The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts execute a secondary validation tool (role-validate) located within the same skill package. The execution uses array-based parameters without a shell, which is a secure practice that prevents command injection.\n- [DATA_EXPOSURE]: The skill reads 1C project files (XML, JSON, and binary metadata) and project configuration (.v8-project.json). This access is required for its documented purpose of managing role permissions and RLS conditions.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from JSON definition files and external text files. Ingestion points include the DefinitionFile parameter and @path values. While specific boundary markers are absent, the skill inventory includes Read, Write, and Bash capabilities. The skill performs sanitization by escaping XML special characters using dedicated functions (Esc-XmlText in PowerShell and lxml in Python) before data interpolation.

role-edit