tartarusai-cli-coding-agent
Fail
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides installation instructions that download executable binaries from a non-trusted GitHub repository (github.com/Tartarus-AI/tartarusai-cli) for both Linux and Windows systems.
- [COMMAND_EXECUTION]: Instructions recommend running the downloaded binary with executable permissions (chmod +x) and moving it to a system binary path (/usr/local/bin/) using elevated privileges (sudo), which is a high-risk operation for untrusted software.
- [REMOTE_CODE_EXECUTION]: By executing an opaque binary downloaded from an untrusted remote source, the skill effectively enables arbitrary code execution on the host machine.
- [PROMPT_INJECTION]: The skill directs the agent to ingest large amounts of local source code (up to 256K tokens) as context. This design lacks boundary markers or sanitization, creating a surface for indirect prompt injection where malicious instructions embedded in a codebase could cause the agent to perform unauthorized tasks such as exfiltrating the stored CLI token or credentials.
Recommendations
- AI detected serious security threats
Audit Metadata