The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from a LeanIX workspace (fact sheet names, attributes, and metadata) via tools like mcp__leanix__search_fact_sheet_by_name and mcp__leanix__text_to_fact_sheets. This data is then interpolated into diagram elements created through other MCP tools.
[INDIRECT_PROMPT_INJECTION]: Golden Rule 4 explicitly instructs the agent to pass HTML markup literally (raw) when using styled text containers with html=1 (e.g., <b>SAP</b>, <br>). This creates a Cross-Site Scripting (XSS) surface where malicious content stored in the LeanIX workspace could be rendered as active scripts or elements when a user views the generated diagram.
[INDIRECT_PROMPT_INJECTION]: The skill lacks instructions for sanitizing or validating external data retrieved from workspace tools before placing it into the diagram's value or label attributes, relying on the underlying platform's handling of the XML which is explicitly bypassed for HTML content.

compose-diagram