svg-hand-drawn-preview
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests SVG data from external sources (local files or URLs) and incorporates this content into a generated
preview.htmlfile using the{{SVG_MARKUP}}placeholder in the provided template. This represents a vulnerability surface where a malicious SVG could contain embedded scripts or event handlers that execute in the user's browser context upon opening the preview. - Ingestion points: The skill explicitly accepts local SVG file paths or remote SVG URLs as input (SKILL.md).
- Boundary markers: There are no specific instructions to use delimiters or sanitization routines to isolate the untrusted SVG data from the generated HTML/JS execution environment.
- Capability inventory: The skill utilizes the agent's file-system capabilities to write multiple output files (
preview.html,player.js,animated.svg,animation-plan.json). - Sanitization: While the workflow suggests inspecting the SVG for unsupported constructs (like
foreignObjector masks), it does not explicitly mandate the stripping of<script>tags or other active content that could lead to XSS. - [COMMAND_EXECUTION]: The skill instructions direct the agent to generate several files directly for the user as the default deliverable. This involves the execution of file-writing commands across the current working directory, which is appropriate for the skill's primary purpose but represents a significant capability level.
Audit Metadata