The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external Office Open XML (.docx) files, which represents an attack surface for indirect prompt injection if the document content contains malicious instructions intended to manipulate the agent's behavior.
Ingestion points: The skill parses content from various XML components within the document package, including word/document.xml (body text), word/styles.xml, and metadata fields like docProps/core.xml (Title, Author).
Boundary markers: The instructions do not define specific delimiters or "ignore" instructions for the content being processed.
Capability inventory: The skill identifies as a "read-only scanner" and produces structured findings. It does not contain instructions for tool use involving file-writing, network exfiltration, or command execution.
Sanitization: No explicit sanitization or validation of the text content extracted from the XML elements is described.
[EXTERNAL_DOWNLOADS]: The skill contains references to official documentation from trusted organizations.
Evidence: The documentation points to authoritative resources hosted on the w3.org and microsoft.com domains, such as the WCAG 2.2 specification and Microsoft Word accessibility support pages.

word-accessibility