The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automates 1C metadata management by executing local scripts .claude/skills/template-add/scripts/add-template.ps1 and add-template.py. These scripts create directory structures and modify existing project XML files based on the provided object and template names.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection or data corruption due to the unsanitized interpolation of user-provided arguments into XML templates.
Ingestion points: ObjectName, TemplateName, and Synonym parameters defined in SKILL.md and passed to the execution scripts.
Boundary markers: No delimiters or safety instructions are used to isolate user-provided strings within the generated XML content.
Capability inventory: The scripts perform directory creation (New-Item, os.makedirs) and file write operations ([System.IO.File]::WriteAllText, open().write()) to modify the local source repository.
Sanitization: The implementation in both PowerShell and Python directly inserts user strings into XML blocks (e.g., <v8:content>$Synonym</v8:content>) without performing XML entity escaping, which could lead to malformed files if special characters are used.

template-add