skills/lexmount/moli/moli-websearch/Gen Agent Trust Hub

moli-websearch

Fail

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides explicit instructions for the AI to install the Moli CLI tool if it is not found on the system. This involves fetching scripts from GitHub and piping them directly to shell interpreters (sh on Linux/macOS and iex on Windows), allowing for the execution of arbitrary remote code.\n- [PRIVILEGE_ESCALATION]: The Windows installation command uses -ExecutionPolicy ByPass, which is a deliberate attempt to circumvent local security settings that restrict script execution. This allows the agent to execute code that might otherwise be blocked by system policies.\n- [COMMAND_EXECUTION]: The skill uses several local execution methods to perform its functions, including moli fetch, moli serve, curl, and a Node.js driver script (reverse-image.cjs) that uses execFileSync to communicate with local services.\n- [EXTERNAL_DOWNLOADS]: The skill fetches installation scripts and potentially other resources from github.com/lexmount/moli. While targeting a public repository, the automated execution of these downloads poses a significant supply chain risk.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to search and retrieve content from a wide variety of third-party search engines and websites, providing a surface for malicious content to influence the agent. 1. Ingestion points: Web search results and image search result pages (SKILL.md, websearch.md). 2. Boundary markers: Absent; fetched content is dumped as Markdown directly into the context. 3. Capability inventory: Shell command execution (moli, curl, node) and network access. 4. Sanitization: Absent; the skill uses raw Markdown conversion of untrusted web content.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Nov 14, 2026, 11:01 AM