OASB Skills Security Benchmark

Open Agent Security Benchmark results for AI-agent skills security scanners, measuring precision, recall, F1, false-positive rate, flag rate, and scan latency.

3rows
f1primary metric
2026-05-27sampled

Metadata

ID
oasb_skills
Category
Security
Release
Unknown

Metrics

Precision, Recall, F1, Flag Rate, False Positive Rate (lower is better), Average Scan Time (lower is better)

Latest Results

Rows parsed from OASB's public benchmark-results-v5.json. The benchmark reports skills-security scanner precision, recall, F1, false-positive rate, and scan-time metrics.

Rank Subject F1 Model Match Provenance Sampled
1 NanoMind TME v0.5.0 (model only) 0.892 — Imported 2026-05-27
2 HMA Full Pipeline (AST + NanoMind v0.5.0) 0.813 — Imported 2026-05-27
3 HMA Static Patterns (no NanoMind) 0.675 — Imported 2026-05-27