Every claim on this site is backed by a reproducible test. Run them yourself with npm test and python -m pytest tests/. The test suite is open source and runs in CI on every commit.
Tests are organized by risk level, not just by module. Higher tiers are harder to pass.
Does it load? Does it return the right type? Fast, high-coverage, catches regressions immediately.
Individual functions tested in isolation. Covers edge cases, boundary values, and type contracts.
Tests that cross module boundaries. Ensures the 14-layer pipeline flows data correctly between layers.
100+ iterations of randomized inputs using fast-check (TS) and Hypothesis (Python). If the property holds under 100 random inputs, it probably holds.
ML-KEM-768, ML-DSA-65, AES-256-GCM boundary tests. Ensures PQC primitives reject malformed inputs and maintain key isolation.
Red-team style tests: prompt injection, cascade attacks, domain drift, replay attacks, timing attacks. The governance gate must DENY or QUARANTINE every one.
| Test files | 174 |
| Tests | 5,957 passed, 8 skipped |
| Runtime | 65 seconds |
| Coverage target | 80% lines/functions, 70% branches |
| Property tests | 100+ iterations per property |
| Tests | 785 passed |
| Markers | enterprise, homebrew, quantum, security, api, crypto |
| Async | asyncio_mode = auto |
| Coverage target | 80% lines/functions, 70% branches |
20+ test files covering L1 through L14. Each layer's input/output contract, axiom compliance, and cross-layer data flow are verified independently.
ML-KEM-768 key encapsulation, ML-DSA-65 digital signatures, AES-256-GCM symmetric encryption. 100% pass rate on all crypto modules.
ALLOW / QUARANTINE / ESCALATE / DENY decision paths tested with known-good and known-bad inputs. Zero false positives on benign inputs.
6 constructed languages (KO/AV/RU/CA/UM/DR), 256 tokens each, phi-weighted. Bijective encoding/decoding verified across all 1,536 tokens.
HYDRA swarm coordination, consensus protocols, ledger integrity, browser agent sandboxing. Tests cover both cooperative and adversarial multi-agent scenarios.
TypeScript and Python implementations tested for identical outputs on the same inputs. Ensures the canonical (TS) and reference (Python) implementations agree.
53 GitHub Actions workflows run on every push and PR. The gate doesn't open unless all tests pass.
| ci.yml | Build + test + lint |
| scbe-gates.yml | SCBE-specific security gates |
| security-checks.yml | Vulnerability scanning |
| conflict-marker-guard.yml | Merge conflict detection |
| weekly-security-audit.yml | Automated weekly audit |
| nightly-connector-health.yml | Connector health checks |
| auto-changelog.yml | Changelog generation |
| huggingface-sync.yml | Model sync to HF |