Analytics

ARS trends, attack-category radar, and regression history. Chat with AnalyticsBot to pivot the view.

Average ARS

Peak ARS

Trend (first → last)

-36

Runs analyzed

ARS over time · gateway

Last 30 days

v1.4.0

v1.3.8

v1.3.6

v1.3.4

v1.2.9

v1.2.0

Attack radar · current vs previous

CurrentPrevious

Category trend

Prompt injection

100%

95%

100%

Data exfiltration

94%

100%

97%

72%

100%

95%

Excessive agency

100%

90%

60%

100%

86%

SOC 2 gates

84%

90%

88%

65%

90%

79%

Run timeline

v1.4.0ARS 74blocked

release/v1.4 · Today

226 pass · 16 warn · 5 fail

2 critical — tool scope + output injection

v1.3.8ARS 22deployed

hotfix/rate-limit · 2d ago

247 pass · 0 warn · 0 fail

Hotfix — auto-approved, clean run

v1.3.6ARS 55deployed

release/v1.3 · 9d ago

238 pass · 9 warn · 0 fail

Security reviewed and approved

v1.3.4ARS 81denied

release/v1.3-rc2 · 17d ago

187 pass · 38 warn · 22 fail

Denied — 3 critical, excessive agency failures

v1.2.9ARS 61exception

release/v1.2 · 28d ago

230 pass · 17 warn · 0 fail

Exception approved — feature flag mitigation

v1.2.0ARS 38deployed

release/v1.2-base · 45d ago

244 pass · 3 warn · 0 fail

Auto-approved

AnalyticsBotcontext: analytics

Ask for pivots, regression windows, or anomaly scans.

Which asset on agentops-hub should we fix first, and what can we deprioritize?

Fix first: document-retrieval-tool — ARS 81 (CRITICAL). It carries a critical indirect-prompt-injection finding (exploitability 84) and fans out to 14 downstream assets (blast 97). Deprioritize openai-agent-loop — ARS 41: exposed (72) but no exploit path today (exploitability 0). Ranked by your blast radius, not raw CVSS. Want the contributor breakdown for document-retrieval-tool?