prompt-engineering-output-xml
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The file contains instructional text referencing common injection phrases like 'Ignore all previous instructions'; however, these are used strictly as illustrative examples of malicious input within a defensive context to show how XML sandboxing can neutralize such attempts.
- [INDIRECT_PROMPT_INJECTION]: The skill provides templates designed for ingesting untrusted external data, which presents an attack surface for indirect injection. * Ingestion points: Untrusted data is processed via placeholders such as {{user_message}} in the provided templates in SKILL.md. * Boundary markers: The skill explicitly promotes the use of XML tags (e.g., <user_input>) as delimited regions to separate data from instructions. * Capability inventory: The skill is configured with access to file system tools including Read, Edit, Write, Glob, and Grep. * Sanitization: The documentation recommends techniques for sandboxing and character escaping, while acknowledging these are not absolute security boundaries.
Audit Metadata