excalidraw
Pass
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill's primary function is architectural: optimizing agent context by offloading the parsing of high-token Excalidraw metadata to isolated subagents.
- [PROMPT_INJECTION]: The skill uses strong, imperative language (e.g., "Iron Law", "NEVER", "STOP") to enforce architectural delegation. This is assessed as non-malicious steering intended for performance optimization and context management rather than bypassing safety protocols.
- [PROMPT_INJECTION]: The skill defines a surface for Indirect Prompt Injection through the processing of diagram files.
- Ingestion points: Files with extensions
.excalidrawand.excalidraw.json(referenced in SKILL.md). - Boundary markers: The provided task templates for subagents lack explicit delimiters or instructions to ignore embedded commands within the JSON text elements.
- Capability inventory: The delegation logic grants subagents the ability to read and write files (via standard agent tools) to modify or analyze diagrams.
- Sanitization: There are no instructions for validating the structure of the JSON or sanitizing the text labels extracted from the files.
Audit Metadata