The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.
Rust agent-orchestrationai-agentsai-assistantdesktopllmlocal-firstmcppersonal-aiprivacyrustsecond-braintauri
test(agent_harness_e2e): restrict visible tools in boundary fixture
Steven Enamakela8ff7922ce83
diff --git a/tests/agent_harness_e2e.rs b/tests/agent_harness_e2e.rsindex 6039504269..e30a6f693b 100644--- a/tests/agent_harness_e2e.rs+++ b/tests/agent_harness_e2e.rs@@ -4623,10 +4623,9 @@ mod tool_policy_boundary_placement {use super::streaming_support::ScriptedProvider;use tinyinference_llm::model::{ChatModel, ModelProfile};- /// Two tools at different permission levels. A `read_only` channel- /// permission blocks the write one, and that restriction is what makes the- /// boundary render at all — `render_tool_policy_boundary` returns `None`- /// unless `session.has_restrictions()` (`tools/agent_policy/prompt.rs:12`).+ /// Two tools at different permission levels. The fixture exposes only the+ /// read tool, which makes the write tool restricted and causes the policy+ /// boundary to render.struct ScopedTool {name: &'static str,level: PermissionLevel,@@ -4676,6 +4675,7 @@ mod tool_policy_boundary_placement {level: PermissionLevel::Write,}),])+ .visible_tool_names(std::collections::HashSet::from(["boundary_read".to_string()])).tool_dispatcher(Box::new(NativeDialect)).workspace_dir(workspace.path().to_path_buf()).event_context("boundary-session", "internal")