Changes to sandbox command assessment feature based on initial experiment feedback (#6091)

* Removed sandbox risk categories; feedback indicates that these are not
that useful and "less is more"
* Tweaked the assessment prompt to generate terser answers
* Fixed bug in orchestrator that prevents this feature from being
exposed in the extension

codex-rs/protocol/src/approvals.rs

@@ -16,24 +16,10 @@ pub enum SandboxRiskLevel {
     High,
 }
 
-#[derive(Debug, Clone, Copy, Deserialize, Serialize, PartialEq, Eq, Hash, JsonSchema, TS)]
-#[serde(rename_all = "snake_case")]
-pub enum SandboxRiskCategory {
-    DataDeletion,
-    DataExfiltration,
-    PrivilegeEscalation,
-    SystemModification,
-    NetworkAccess,
-    ResourceExhaustion,
-    Compliance,
-}
-
 #[derive(Debug, Clone, Deserialize, Serialize, PartialEq, Eq, JsonSchema, TS)]
 pub struct SandboxCommandAssessment {
     pub description: String,
     pub risk_level: SandboxRiskLevel,
-    #[serde(default, skip_serializing_if = "Vec::is_empty")]
-    pub risk_categories: Vec<SandboxRiskCategory>,
 }
 
 impl SandboxRiskLevel {
@@ -46,20 +32,6 @@ impl SandboxRiskLevel {
     }
 }
 
-impl SandboxRiskCategory {
-    pub fn as_str(&self) -> &'static str {
-        match self {
-            Self::DataDeletion => "data_deletion",
-            Self::DataExfiltration => "data_exfiltration",
-            Self::PrivilegeEscalation => "privilege_escalation",
-            Self::SystemModification => "system_modification",
-            Self::NetworkAccess => "network_access",
-            Self::ResourceExhaustion => "resource_exhaustion",
-            Self::Compliance => "compliance",
-        }
-    }
-}
-
 #[derive(Debug, Clone, Deserialize, Serialize, JsonSchema, TS)]
 pub struct ExecApprovalRequestEvent {
     /// Identifier for the associated exec call, if available.

codex-rs/protocol/src/protocol.rs

@@ -37,7 +37,6 @@ use ts_rs::TS;
 pub use crate::approvals::ApplyPatchApprovalRequestEvent;
 pub use crate::approvals::ExecApprovalRequestEvent;
 pub use crate::approvals::SandboxCommandAssessment;
-pub use crate::approvals::SandboxRiskCategory;
 pub use crate::approvals::SandboxRiskLevel;
 
 /// Open/close tags for special user-input blocks. Used across crates to avoid