penpot-audit-accessibility
Pass
Audited by Gen Agent Trust Hub on Jul 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill strictly follows a read-only architecture. It explicitly instructs the agent to 'Report, don't silently fix' and prohibits canvas mutations, ensuring that the tool acts only as an advisor.
- [COMMAND_EXECUTION]: The skill utilizes the
execute_codetool to run embedded JavaScript files for design analysis. These scripts (e.g.,collectAccessibilityData.js,checkColorContrast.js) process local design data to calculate accessibility metrics. The scripts do not perform any network operations, external downloads, or file system modifications outside of the Penpot environment. - [PROMPT_INJECTION]: The skill processes untrusted content by reading text characters from the design canvas to perform its audit. This represents a potential surface for indirect prompt injection. However, the skill implements a mitigation by truncating the captured text to the first 40 characters, which limits the likelihood of an embedded instruction successfully hijacking the agent's context.
Audit Metadata