word-accessibility
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to ingest and analyze untrusted data from external .docx files.
- Ingestion points: The agent parses content from various XML components of the Word document, such as
word/document.xmlandword/styles.xml(SKILL.md). - Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore instructions that might be embedded within the document text it scans.
- Capability inventory: The skill defines itself as a read-only scanner focused on producing accessibility reports, which limits its potential for harmful actions even if an injection occurs.
- Sanitization: The instructions do not specify any filtering or escaping of content extracted from attributes like 'descr' (alt text) or '<w:t>' (paragraph text).
- [EXTERNAL_DOWNLOADS]: The skill references authoritative documentation from well-known and trusted organizations such as the W3C and Microsoft. These links point to official specifications and support pages for WCAG compliance and Word document accessibility.
Audit Metadata