z-skill-mastery
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs authors on how to handle untrusted source material using XML delimiters in
references/xml-patterns.md. While it correctly notes that delimiters do not establish a security boundary, it lacks specific instructions for sanitization or escaping of the content placed within these tags. - Ingestion points: The skill explicitly suggests patterns for wrapping untrusted source text (e.g.,
<source> [Untrusted source text] </source>). - Boundary markers: It uses XML tags as markers but explicitly warns that they 'do not make that data trusted or establish a security boundary'.
- Capability inventory: The skill itself is instructional (Markdown-based) and does not invoke subprocesses or network operations, but it guides the creation of skills that may have such capabilities.
- Sanitization: No specific sanitization or escaping mechanisms (like escaping
</source>tags within the input) are provided in the templates.
Audit Metadata