diff --git a/plugin-catalog/storm-fusion-research.yaml b/plugin-catalog/storm-fusion-research.yaml
index 6ff4277a34..e306615fa4 100644
--- a/plugin-catalog/storm-fusion-research.yaml
+++ b/plugin-catalog/storm-fusion-research.yaml
@@ -1,7 +1,7 @@
 name: storm-fusion-research
 repo: https://github.com/bmtrnavsky/storm-fusion-research-pipeline
 sha: 76e322bede3d039632672b833422f0146dce15d6
-description: Human-gated perspective-panel research. One question, several seats, one judge, one human. Solo, native subagent panel, or per-seat MoA/Fusion fanout via the storm_run_panel tool.
+description: "Human-gated perspective-panel research. One question, several seats, one judge, one human. Solo, native subagent panel, or per-seat MoA/Fusion fanout via the storm_run_panel tool. Disclosure — every seat/judge call runs through Hermes's plugin LLM lane on your own credentials; with llm.allow_provider_override / allow_model_override enabled the agent picks provider and model per seat and spend follows that choice. The human gates are skill instructions to the agent, not enforced by the tool."
 maintainer: bmtrnavsky
 tier: community
 category: tools