Purpose and security role
Cerberus positions itself as a defensive guardian system for prompt injection, jailbreak, and input-threat surfaces, with a central hub that coordinates multiple guardian strategies.
Guardian and hub structure
- HubCoordinator as central decision aggregator
- Current guardian types: Strict, Pattern, Heuristic, Statistical
- Direction includes rate-limiting, monitoring, access-control, and audit logging
Maturity and test posture
- Verified Package metadata: version 0.1.0, Development Status 2 - Pre-Alpha.
- Documented README advertises 104 tests and 59% coverage.
- Partial Production-grade wording in README is stronger than pre-alpha classifier.
Validation and threat surfaces
Documented input-validation and threat-detection surfaces include SQLi/XSS/path traversal signatures, behavioral heuristics, and anomaly detection, with planned or directional modules for monitoring and compliance.
Cross-platform statement is documented for Linux/macOS/Windows with noted Windows sandbox limitations; this portal has not independently reproduced those runtime claims.
Thirsty-Lang binding status is explicitly documented in README as 0.8.3 API binding language and should not be promoted to broader compatibility claims without newer source proof.
Limitations and proof trail
- Partial Production-readiness language conflicts with pre-alpha metadata.
- Unverified Test/coverage values were not re-run from this website update task.
- Documented Threat surfaces are traced to pinned public docs.