drawio-diagram-forge
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input data (text, images, and Excel) to generate structured XML diagrams, creating a surface for potential instruction injection via data processed at runtime.
- Ingestion points: Arbitrary text, images, or Excel files provided by the user as described in
SKILL.md. - Boundary markers: The prompt instructions lack explicit delimitation or 'ignore embedded instructions' warnings for the data processing phase.
- Capability inventory: The skill allows the agent to write files (
.drawio,.drawio.svg) and execute local Python validation scripts (scripts/validate_drawio.py). It also references usage of thedraw.ioCLI for exports. - Sanitization: While
references/troubleshooting.mddiscusses HTML encoding to prevent rendering issues, there is no security-focused sanitization or filtering of external content described to prevent instructions embedded in inputs from influencing the agent's behavior.
Audit Metadata