moli-cdp-server
Warn
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill directs the agent to download and pipe shell scripts directly into an interpreter (sh or PowerShell). Evidence:
curl ... | shandirm ... | iexin SKILL.md. - [EXTERNAL_DOWNLOADS]: The skill downloads installer scripts from the author's GitHub repository (
github.com/lexmount/moli). While the source is the vendor, the automated download-and-execute pattern is a security risk. - [PRIVILEGE_ESCALATION]: The installation instructions for Windows include the
-ExecutionPolicy ByPassflag, which allows the execution of scripts that would otherwise be blocked by system security settings. - [COMMAND_EXECUTION]: The agent is instructed to run the
molibinary and manage themoli serveprocess. - [INDIRECT_PROMPT_INJECTION]: The skill enables web browsing via Playwright/Puppeteer. 1. Ingestion points: External websites browsed via CDP. 2. Boundary markers: Absent. 3. Capability inventory: Execution of local binaries and network server management. 4. Sanitization: Absent. This provides a vector for untrusted web content to inject instructions into the agent's context.
Audit Metadata