| Version | Supported |
|---|---|
| 0.18.x | Active |
| 0.17.x | Security fixes |
| < 0.17 | No longer supported |
Do not open a public issue. Email: w_shixiang@163.com
Response within 48 hours. Fixes published within 7 days.
#![forbid(unsafe_code)]— No unsafe Rust in core and web crates- Shell injection prevention — Commands parsed and sanitized before execution
- Path traversal protection — I/O paths validated against workspace boundaries
- API key encryption — Provider keys encrypted at rest
- Rate limiting — Per-IP on auth and chat endpoints
- No secrets in source — Credentials via env vars or encrypted config
- Input validation — Structured error responses
{code, message, detail}
AI service layer: zero DB write, zero process management, zero filesystem write. AI never executes autonomously — every action requires human approval.
cargo audit runs in CI. Cargo.lock committed. Vulnerabilities addressed within 7 days.