jfrog
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill checks for the latest available version of the JFrog CLI by querying the official vendor release repository at
releases.jfrog.io. - [REMOTE_CODE_EXECUTION]: Instructions are provided to install or upgrade the JFrog CLI using a shell script fetched from the official vendor domain
install-cli.jfrog.io. While this uses a high-risk pipe-to-shell pattern, it targets a trusted vendor resource. - [COMMAND_EXECUTION]: The skill executes the
jfCLI and internal Node.js/Bash scripts (e.g.,check-environment.mjs,jfrog-login-save-credentials.sh) to manage authentication, environment state, and platform tasks. - [DATA_EXFILTRATION]: The environment check scripts construct a
JFROG_CLI_USER_AGENTstring containing metadata such as the agent name, the hosting IDE (e.g., Cursor, VS Code), and the AI model slug. this metadata is sent to JFrog's infrastructure via standard HTTP headers during authenticated requests for telemetry and support purposes. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external JFrog API responses. To mitigate risks, the instructions strictly mandate the use of
jqfor structured parsing and the use of private temporary directories (JF_TMP) to isolate and preserve command output.
Audit Metadata