The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local PowerShell and Python scripts (skd-info.ps1, skd-info.py) to parse 1C:Enterprise metadata and report templates. These scripts resolve file paths and use XML parsing libraries to extract schema information.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted XML data from 1C report templates (Template.xml), which often contains complex queries and descriptive text that is returned to the agent's active context.
Ingestion points: Data enters the context via the Template.xml file and related metadata files found in the project directory (e.g., Ext/Template.xml).
Boundary markers: The skill does not use specific delimiters or protective headers in its output to warn the agent that the returned text contains potentially untrusted content from the schema.
Capability inventory: The skill has access to the local file system through Read, Glob, and Bash tools, and it executes scripts that can read any file reachable by the agent process.
Sanitization: There is no evidence of sanitization or escaping of the content extracted from XML fields (such as queries or titles) before it is passed to the agent.

skd-info