Benchmarks & Engineering Validation
This page reports HunterX v7 engineering validation from the v7 certification and hardening reports. Figures are engineering metrics — not marketing claims and not guarantees of discovery or detection rates.
Test suite
At the v7.0.0 release (Sprint 035 final hardening audit):
| Result | Count |
|---|---|
| Passed | 3479 |
| Skipped | 8 |
| Deselected | 2 |
| Failed | 0 |
The suite spans unit, component, integration, golden, security, acceptance, performance, engineering, architecture and framework tests, plus the 92-tool toolchain contract suite.
Quality gates
At release, the following gates were green: pytest, ruff, mypy (eng + shared), bandit (Medium+), vulture, docs (7/7) and lock-file consistency.
Known engineering follow-ups (P2, non-blocking). The v7 certification reports record a performance-gate self-fail issue: the performance gate flags its own benchmark tests (>20s) as slow and cannot pass as configured, even though measured throughput is fine. Coverage gate reports 77% branch-adjusted vs an 80% line-rate target. These are internal quality-gate tuning items, not claims about HunterX detection performance.
Measured performance
Measured performance figures are reported in the v7 engineering certification
reports (docs/v7-sprint-034-final-engineering-certification.md) from the
benchmark suite (tests/performance). Representative figures include
throughput on the order of >100k ops/s for correlation, >2.5M ops/s for DNS
resolution and >8M ops/s for confidence computation. Performance tests cover
mission scale (1/10/100 missions), observation scale (10k/100k/1M) and
large finding/evidence sets.
These are internal engineering measurements from the certification environment. Real-world performance depends on your environment, target, scope and tool selection.
Interpreting these numbers
HunterX is an AI-assisted discovery, validation and proof engine, not a race-to-fastest scanner. Its differentiator is verified findings with evidence, reproducibility and impact — not raw detection speed. Use the figures above as engineering context, and validate against your own workloads.