agentar
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads an installation shell script and a package tarball from
https://dtaiagtavtr-test.dl.alipaydev.com. These resources are hosted on a development domain belonging to a well-known technology service. - [REMOTE_CODE_EXECUTION]: The workflow involves executing a downloaded shell script (
install-agentar-cli.sh) to install theagentar-clitool on the system. This is the primary and intended function of the skill. - [COMMAND_EXECUTION]: The skill utilizes the
agentarCLI for various tasks, including local authentication, chat interactions, and session management. These commands interact with the tool's backend to provide its core functionality. - [PROMPT_INJECTION]: The skill processes natural language input through the CLI (e.g.,
agentar "hello"), which serves as an ingestion point for user-supplied data. While this represents a surface for indirect prompt injection, the skill does not exhibit any patterns of malicious exploitation or safety boundary bypass.
Audit Metadata