feat: add provenance-aware worker history

crates/memory/src/extract/payload.rs

@@ -9,7 +9,7 @@
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
-use crate::schema::{EvidenceKind, SourceEvidenceRef, SourceRef};
+use crate::schema::{EvidenceKind, EvidenceOrigin, SourceEvidenceRef, SourceRef};
 
 /// Current flat staging schema version.
 pub const STAGING_SCHEMA_VERSION: u32 = 2;
@@ -80,6 +80,8 @@ pub struct StagingEvidence {
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub entry_range: Option<[u64; 2]>,
     #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub origin: Option<EvidenceOrigin>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
     pub excerpt: Option<String>,
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub summary: Option<String>,
@@ -159,6 +161,7 @@ mod tests {
             id: "E001".into(),
             kind: EvidenceKind::new(EvidenceKind::MESSAGE),
             entry_range: Some([10, 12]),
+            origin: None,
             excerpt: Some("extract candidate taxonomy".into()),
             summary: Some("User and assistant discussed staging kinds".into()),
         };

crates/memory/src/schema/common.rs

@@ -67,6 +67,40 @@ impl EvidenceKind {
     }
 }
 
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum EvidenceOriginKind {
+    HumanInput,
+    WorkerInput,
+    FlowInstruction,
+    BackendInstruction,
+    ModelOutput,
+    ToolOutput,
+    DerivedSummary,
+    LegacyUnknown,
+}
+
+/// Bounded origin snapshot attached to extraction evidence. This is audit
+/// metadata only and cannot authorize Workspace operations.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+pub struct EvidenceOrigin {
+    pub kind: EvidenceOriginKind,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub account_id: Option<String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub workspace_id: Option<String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub runtime_id: Option<String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub worker_id: Option<String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub flow_selector: Option<String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub flow_definition_id: Option<String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub flow_definition_revision: Option<u64>,
+}
+
 /// Host-resolved source/evidence metadata for an individual staging claim.
 ///
 /// This deliberately stores only bounded anchor metadata: stable ids, entry
@@ -86,6 +120,9 @@ pub struct SourceEvidenceRef {
     /// Host-assigned evidence id within the referenced evidence set.
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub evidence_id: Option<String>,
+    /// Trusted typed origin snapshot for this logical evidence entry.
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub origin: Option<EvidenceOrigin>,
     /// Extensible evidence kind tag.
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub evidence_kind: Option<EvidenceKind>,

crates/memory/src/schema/mod.rs

@@ -10,7 +10,10 @@ mod decision;
 mod request;
 mod summary;
 
-pub use common::{EvidenceKind, Frontmatter, SourceEvidenceRef, SourceRef, split_frontmatter};
+pub use common::{
+    EvidenceKind, EvidenceOrigin, EvidenceOriginKind, Frontmatter, SourceEvidenceRef, SourceRef,
+    split_frontmatter,
+};
 pub use decision::{DecisionFrontmatter, DecisionStatus};
 pub use request::RequestFrontmatter;
 pub use summary::SummaryFrontmatter;