claude-better-cli
Fail
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs users to obtain and execute a binary from an external, third-party GitHub repository (
github.com/krzyzanowskim/claude-better). The content explicitly states that source code is restricted ('provided for selected high-profile customers only'), which prevents users and the AI agent from verifying the security or integrity of the binary before execution. - [COMMAND_EXECUTION]: The installation guide promotes hijacking the standard
claudecommand by modifying the systemPATHor using analias. This pattern is a common technique for intercepting calls to legitimate software to perform malicious actions in the background. - [CREDENTIALS_UNSAFE]: As a 'drop-in replacement' for the official CLI, the tool expects the user to provide their
ANTHROPIC_API_KEYand allows it access to the~/.claude/configuration directory. Granting access to these sensitive credentials to an unverified third-party implementation poses a high risk of credential exfiltration. - [EXTERNAL_DOWNLOADS]: The skill directs users to a third-party website (
ara.so) and an untrusted GitHub repository to download execution components. These sources do not belong to established trusted organizations.
Recommendations
- AI detected serious security threats
Audit Metadata