design-taste-frontend

Fail

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: Automated security scanners have confirmed that the file library/sites.json contains malware. Furthermore, the skill's dataset includes several blacklisted domains, such as https://thedeepend.space and https://extension.textmode.art. These findings indicate that the skill is configured to interact with known malicious resources.
  • [EXTERNAL_DOWNLOADS]: The scripts/assets.mjs script allows the agent to search for and download image and video files from external sources (Openverse, Wikimedia Commons, NASA, etc.) directly into the project's local filesystem.
  • [DYNAMIC_EXECUTION]: The skill uses the Chrome DevTools Protocol to control a web browser. In scripts/inspect.mjs, it uses Runtime.evaluate to execute dynamically generated JavaScript strings inside the browser, providing a mechanism for potential script injection.
  • [COMMAND_EXECUTION]: Browser instances (Chrome, Edge, or Chromium) are launched using child_process.spawn in scripts/lib/cdp.mjs. On Windows systems, the skill also invokes the taskkill command via spawnSync to manage browser lifecycle.
  • [PROMPT_INJECTION]: Instructions in SKILL.md attempt to override default agent behaviors with directives such as "Do not argue with the meter" and "Do not report until all of these are true," which are intended to enforce strict compliance with the skill's internal design rules regardless of standard safety protocols.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted metadata from external media APIs via scripts/assets.mjs. This data is subsequently saved locally and processed, creating an attack surface where an attacker could embed malicious instructions in media metadata to influence the agent.
Recommendations
  • CRITICAL: 1 file(s) identified as malware by FileRep - DO NOT USE
  • AI detected serious security threats
  • Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Nov 11, 2026, 06:53 PM