Skip to content

The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.

docs(util): name tinytools::sanitize as the sanitization owner

M3gA-Mind1946a973a7b8
diff --git a/crates/openhuman-core/src/util/README.md b/crates/openhuman-core/src/util/README.md
index 4507483826..0b9504cb54 100644
--- a/crates/openhuman-core/src/util/README.md
+++ b/crates/openhuman-core/src/util/README.md
@@ -105,9 +105,10 @@ wording stays one contract.
### Sanitization
The stripping rule for untrusted text going into a model's context lives in
-`tinymcp_bus`, which applies it to every remote MCP tool description. This
-module re-exports it at the old path so the orchestrator prompt builder's
-skill descriptions get the same rule. Change it in `vendor/tinymcp`, not here.
+`tinytools::sanitize`, which TinyMCP also applies to every remote MCP tool
+description. The orchestrator prompt builder and skill search call
+`tinytools::sanitize::sanitize_for_llm` directly, so skill descriptions get the
+same rule. Change it in `vendor/tinyagents/vendor/tinytools`, not here.
### `MaybeSet`