From d891cdce22d359095fd529bb937cdfe67f30cff6 Mon Sep 17 00:00:00 2001 From: Greg Anderson Date: Tue, 22 Sep 2026 13:58:43 -0600 Subject: [PATCH] docs(sensei): note how the Advisor's cross-tool estimate reads Dedupe Pools The cross-tool estimate now compares findings within each Asset's cross-tool pool and keeps engagement-scoped deduplication on its own. Say so on the Sensei Advisor page, and that the model sees pool counts, never pool names. Co-Authored-By: Claude Opus 5.5 --- docs/content/sensei/sensei_advisor.md | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/docs/content/sensei/sensei_advisor.md b/docs/content/sensei/sensei_advisor.md index eb262bb8400..589001222c4 100644 --- a/docs/content/sensei/sensei_advisor.md +++ b/docs/content/sensei/sensei_advisor.md @@ -52,6 +52,12 @@ merge, and which tool pairs they fall between. A recommendation whose simulated essentially zero is flagged rather than presented as confident, and one that would collapse unrelated findings onto a single hash is dropped before you ever see it. +The cross-tool estimate compares findings the way deduplication will. Findings in Assets that +share a cross-tool [Dedupe Pool](/triage_findings/finding_deduplication/pro__dedupe_pools/) are +compared with each other, and an Engagement set to deduplicate only within itself is compared +with nothing outside it. When pools shaped the estimate, the card says so. The model is told how many pools and pooled Assets exist, never +their names. + Applying a recommendation uses the same settings machinery and permissions as changing the setting by hand — applying a deduplication recommendation requires the Tuner edit permission, and creating a prioritization engine requires the engine permission. The Advisor grants no new