skills/mobbin/skills/redesign-screen/Gen Agent Trust Hub

redesign-screen

Pass

Audited by Gen Agent Trust Hub on Oct 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [SAFE]: The skill follows a well-documented, multi-step design process that prioritizes evidence-based decision-making. All external references are restricted to the vendor's own Mobbin service, and the use of internal templates for rendering visual outputs is consistent with its stated purpose as a design tool.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface as it ingests untrusted data from user-provided 'sources' (such as codebases, design files, or images) and external Mobbin search results. These inputs are processed and used to generate design plans and rendered HTML/PNG boards.
  • Ingestion points: User-provided codebase files, design tool data, and metadata from Mobbin search results (app names, URLs).
  • Boundary markers: The workflow includes a 'Freeze the baseline' step and 'Ground rules' that logically separate user input from reference data, although no explicit technical sanitization (like escaping) is mentioned for ingested code symbols.
  • Capability inventory: The skill possesses file-system read/write capabilities (editing source code in 'Targeted' mode) and tool-based network access for design pattern retrieval.
  • Sanitization: Instructions do not explicitly detail the sanitization of external content before it is interpolated into the generated reports or HTML templates. However, this risk is inherent to the tool's primary function and is mitigated by the agent's own safety guardrails and the structured nature of the workflow.
  • [DYNAMIC_EXECUTION]: The skill utilizes local HTML templates (assets/board.html, assets/variations.html) that employ JavaScript to dynamically render user interfaces. The templates use innerHTML to inject content generated by the agent (e.g., move descriptions, references). While the use of innerHTML can be a vector for script injection if data is malicious, in this context, the data is primarily generated by the agent based on its analysis, and the templates are executed in a sandboxed rendering environment (browser capture tool).
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 14, 2026, 01:58 AM