The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs local script execution to validate XML changes after editing operations.
Evidence: In scripts/subsystem-edit.ps1, it calls powershell.exe -NoProfile -File $validateScript. In scripts/subsystem-edit.py, it uses subprocess.run([sys.executable, validate_script, ...]). The path to the validation script is resolved relative to the skill's installation directory, and it targets a sibling tool (subsystem-validate) from the same author.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from XML and JSON files which could serve as an attack surface for indirect instructions.
Ingestion points: XML subsystem files provided via SubsystemPath, JSON definition files via DefinitionFile, and JSON strings in the Value argument.
Boundary markers: No explicit prompt boundary markers or 'ignore embedded instructions' warnings are present for the data being processed.
Capability inventory: The skill can write to the local file system (WriteAllText and Python's open().write()) and execute local scripts for validation.
Sanitization: The skill implements robust XML escaping via esc_xml and esc_xml_text functions (and .NET's SecurityElement.Escape in PowerShell) to ensure that user-provided values are safely interpolated into the XML structure.
[DATA_EXPOSURE]: The skill accesses project-specific configuration files to enforce development best practices.
Evidence: It reads .v8-project.json, Configuration.xml, and binary files like Ext/ParentConfigurations.bin. This 'support guard' logic is used exclusively to determine if a 1C object is on vendor support and should be protected from direct editing, which is a safety feature for the user.

subsystem-edit