diff --git a/README.md b/README.md
index 31841e59..a2953c93 100644
--- a/README.md
+++ b/README.md
@@ -148,6 +148,27 @@ scripts/
   conways-rules.txt # Core rules for the automaton
 ```
 
+## Audit event fields (risk/policy)
+
+Tool-call results now support optional structured audit metadata to make policy outcomes easier to analyze downstream:
+
+- `riskLevel`: tool risk classification (`safe|caution|dangerous|forbidden`)
+- `policyDecision`: normalized policy outcome (`allow|block|require_confirmation|dry_run`)
+- `policyReason`: short human-readable reason
+- `capability`: capability namespace for the action (e.g. `network.expose_port`, `finance.transfer_credits`)
+
+Example:
+
+```json
+{
+  "name": "transfer_credits",
+  "riskLevel": "dangerous",
+  "policyDecision": "allow",
+  "policyReason": "ALLOWED: All policy checks passed",
+  "capability": "finance.transfer_credits"
+}
+```
+
 ## License
 
 MIT
diff --git a/src/agent/tools.ts b/src/agent/tools.ts
index f32a069e..6a1acff3 100644
--- a/src/agent/tools.ts
+++ b/src/agent/tools.ts
@@ -3261,6 +3261,35 @@ export function toolsToInferenceFormat(
  * Execute a tool call and return the result.
  * Optionally evaluates against the policy engine before execution.
  */
+function mapCapability(tool: AutomatonTool): string {
+  switch (tool.name) {
+    case "transfer_credits":
+      return "finance.transfer_credits";
+    case "x402_fetch":
+      return "finance.x402_payment";
+    case "expose_port":
+      return "network.expose_port";
+    case "remove_port":
+      return "network.remove_port";
+    case "register_domain":
+    case "add_dns_record":
+    case "delete_dns_record":
+      return "domain.manage";
+    case "self_modify_code":
+    case "install_skill":
+    case "uninstall_skill":
+      return "self_mod.modify";
+    default:
+      return `${tool.category}.${tool.name}`;
+  }
+}
+
+function toPolicyDecisionLabel(action: "allow" | "deny" | "quarantine"): "allow" | "block" | "require_confirmation" | "dry_run" {
+  if (action === "deny") return "block";
+  if (action === "quarantine") return "require_confirmation";
+  return "allow";
+}
+
 export async function executeTool(
   toolName: string,
   args: Record<string, unknown>,
@@ -3284,9 +3313,16 @@ export async function executeTool(
       result: "",
       durationMs: 0,
       error: `Unknown tool: ${toolName}`,
+      policyDecision: "block",
+      policyReason: "TOOL_UNKNOWN: not found in registered tool set",
+      capability: `unknown.${toolName}`,
     };
   }
 
+  const capability = mapCapability(tool);
+  let policyDecisionLabel: "allow" | "block" | "require_confirmation" | "dry_run" = "allow";
+  let policyReason = "No policy engine provided; executed legacy allow path";
+
   // Policy evaluation (if engine is provided)
   if (policyEngine && turnContext) {
     const request: PolicyRequest = {
@@ -3297,6 +3333,8 @@ export async function executeTool(
     };
     const decision = policyEngine.evaluate(request);
     policyEngine.logDecision(decision);
+    policyDecisionLabel = toPolicyDecisionLabel(decision.action);
+    policyReason = `${decision.reasonCode}: ${decision.humanMessage}`;
 
     if (decision.action !== "allow") {
       return {
@@ -3306,6 +3344,10 @@ export async function executeTool(
         result: "",
         durationMs: Date.now() - startTime,
         error: `Policy denied: ${decision.reasonCode} — ${decision.humanMessage}`,
+        riskLevel: tool.riskLevel,
+        policyDecision: policyDecisionLabel,
+        policyReason,
+        capability,
       };
     }
   }
@@ -3368,6 +3410,10 @@ export async function executeTool(
       arguments: args,
       result,
       durationMs: Date.now() - startTime,
+      riskLevel: tool.riskLevel,
+      policyDecision: policyDecisionLabel,
+      policyReason,
+      capability,
     };
   } catch (err: any) {
     return {
@@ -3377,6 +3423,10 @@ export async function executeTool(
       result: "",
       durationMs: Date.now() - startTime,
       error: err.message || String(err),
+      riskLevel: tool.riskLevel,
+      policyDecision: policyDecisionLabel,
+      policyReason,
+      capability,
     };
   }
 }
diff --git a/src/types.ts b/src/types.ts
index 4e967561..5cc87158 100644
--- a/src/types.ts
+++ b/src/types.ts
@@ -118,6 +118,11 @@ export interface ToolCallResult {
   result: string;
   durationMs: number;
   error?: string;
+  /** Structured policy/audit metadata (optional for backward compatibility) */
+  riskLevel?: RiskLevel;
+  policyDecision?: "allow" | "block" | "require_confirmation" | "dry_run";
+  policyReason?: string;
+  capability?: string;
 }
 
 export interface TokenUsage {