CodeWhale: run_tests Tool Enables RCE via Malicious Repository Without Approval
Hmbown CodeWhale
Affected versions
| Product | Affected | Fixed in |
|---|---|---|
| deepseek-tui crates.io |
|
|
| deepseek-tui-cli crates.io |
|
|
| deepseek-tui npm |
Hmbown CodeWhale
| Product | Affected | Fixed in |
|---|---|---|
| deepseek-tui crates.io |
|
|
| deepseek-tui-cli crates.io |
|
|
| deepseek-tui npm |
≥ 0.3.0 · < 0.8.23
|
CodeWhale is a DeepSeek + MiMo coding agent in terminal. From 0.3.0 to 0.8.23, the run_tests tool executes cargo test in the workspace with ApprovalRequirement::Auto, meaning it runs without any user approval prompt. cargo test compiles and executes arbitrary code: test binaries, build.rs build scripts, and proc macros. While auto-approving test execution is a deliberate design choice, it creates an inconsistency in the security boundary. However, in a malicious repository, test code can execute arbitrary shell commands, exfiltrate credentials, or establish persistence with zero approval. The attack is amplified by AGENTS.md (auto-loaded into the system prompt), which can instruct the model to run tests proactively at session start. This vulnerability is fixed in 0.8.23.
From the CVE record · GitHub_M
In affected CodeWhale packages, run_tests executes cargo test without a user approval prompt. In a malicious repository, tests, build scripts, or procedural macros can run arbitrary commands when the agent runs tests; an AGENTS.md file can prompt the agent to do so at session start.
Anyone running crates.io deepseek-tui or deepseek-tui-cli, or npm deepseek-tui, at versions >=0.3.0 and <0.8.23 should check their installation.