drawio-logical-diagrams
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions of systems or processes to generate XML diagrams, which presents a surface for indirect prompt injection.\n
- Ingestion points: User requests and system descriptions (described in
SKILL.md).\n - Boundary markers: The instructions do not define specific delimiters to separate user input from agent instructions.\n
- Capability inventory: The skill is configured with access to
Read,Write, andBashtools in its environment frontmatter.\n - Sanitization: The skill provides a validation checklist that includes escaping special characters (e.g.,
<to<) for XML output, which helps mitigate injection risks into the generated file content.
Audit Metadata