show-me

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes conversational context to generate diagrams and interactive widgets. It lacks explicit instructions for sanitizing or escaping user-provided data before it is rendered into visuals, which could allow malicious content within the conversation to influence the rendering environment.
  • Ingestion points: The skill ingests conversation context, including names, accounts, and error messages, to 'anchor in the live example' (SKILL.md).
  • Boundary markers: The skill defines a markdown 'Output Shape' for framing the visual, but does not provide delimiters to separate untrusted context from the visual generation logic.
  • Capability inventory: The skill utilizes visualization tools (e.g., 'show_widget' MCP) and can write HTML or SVG files to a scratchpad to be opened in a browser.
  • Sanitization: Instructions for sanitizing external content are absent.
  • [DYNAMIC_EXECUTION]: The skill instructs the agent to write HTML or SVG files and open them in a browser to visualize concepts. This involves generating and executing code based on potentially untrusted input from the conversation, which presents a risk of Cross-Site Scripting (XSS) if the data contains malicious payloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 11:19 PM
Security Audit — agent-trust-hub — show-me