maple-preview

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses platform-specific commands including open, xdg-open, and start to open the generated asset preview page in the default web browser. This is an intended function to allow local visualization of requested assets.\n- [EXTERNAL_DOWNLOADS]: The preview template fetches image assets and animation data from the vendor-owned domain https://resource-static.msu.io/data/. This represents standard functionality for retrieving MapleStory assets from the vendor infrastructure.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external metadata such as item names and descriptions, which constitutes a surface for indirect prompt injection. \n
  • Ingestion points: Untrusted data enters the context via search results for item names, IDs, and thumbnail URLs as described in the Building the page section of SKILL.md. \n
  • Boundary markers: The data is injected into the HTML template as a JSON-formatted string within a script tag, which is then parsed by the browser at runtime. \n
  • Capability inventory: The skill has the capability to write HTML files to the .maple-preview/ directory and execute shell commands to open those files. \n
  • Sanitization: The HTML template uses JSON.parse() for data ingestion and populates DOM elements using textContent, which effectively prevents simple cross-site scripting (XSS) attacks from malicious asset names.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 03:23 AM
Security Audit — agent-trust-hub — maple-preview