From Probe Outcomes to Audit-Grade Findings

From Probe Outcomes to Audit-Grade Findings A data-flow diagram generated by Archify. 01 / Collection 02 / Separation 03 / Verification 04 / Mapping 05 / Consumers Probe Catalog · 1,092 unique keys · 01 / Collection · attack source Probe Catalog 1,092 unique keys attack source Evaluated Outcomes · 2,000 rows · 02 / Separation · passed + failed Evaluated Outcomes 2,000 rows passed + failed Blocked Probes · 1,757 · 87.8% · 02 / Separation · defense held Blocked Probes 1,757 · 87.8% defense held Findings · 243 · 82 critical · 03 / Verification · evidence Findings 243 · 82 critical evidence AEGIS-J Verdicts · 208 judged · 04 / Mapping · V-FULL 85 AEGIS-J Verdicts 208 judged V-FULL 85 Confidence Scoring · mean 0.90 · 04 / Mapping · false-positive risk Confidence Scoring mean 0.90 false-positive risk Control Tagging · 86 distinct controls · 05 / Consumers · framework_controls Control Tagging 86 distinct controls framework_controls Coverage Map · 15+ frameworks · 05 / Consumers · OWASP 76% · EU AIA 54% Coverage Map 15+ frameworks OWASP 76% · EU AIA 54% Reports + Exports · SARIF · JUnit · web · 05 / Consumers · consumers Reports + Exports SARIF · JUnit · web consumers probe executions 126 scans severity-ranked findings 243 rows passed probes 87.8% held judge evaluation 208 judged confidence + FP risk mean 0.90 tagged findings 94.8% carry tags rollup 86 controls verified findings 38 criticals @ 100% ASR control evidence audit-grade Legend primary data policy / PII async batch data store data flow

Precision over volume

  • • 87.8% of probes were blocked — the residue is what matters
  • • 243 findings, not thousands of heuristic flags
  • • 38 criticals reproduced at 100% attack success rate

Honest verification

  • • 208 of 243 findings carry explicit judge verdicts
  • • Mean internal confidence 0.90 across judged findings
  • • Replay verification exists in the platform; this window did not exercise it

Audit-ready output

  • • 86 distinct control IDs light up across 15+ frameworks
  • • OWASP LLM 76% · NIST AI RMF 61% · EU AI Act 54% · SOC 2 40%
  • • Exports land as SARIF and JUnit with the same tags