diff --git a/packages/paperclip-runner/README.md b/packages/paperclip-runner/README.md
index 5936d1c7b4..f1d28444d8 100644
--- a/packages/paperclip-runner/README.md
+++ b/packages/paperclip-runner/README.md
@@ -66,6 +66,12 @@ effects, audit, retries, conflicts, redaction, continuation, and terminal
 decisions. The production adapter stays App-owned and invokes Paperclip's real
 route/service authorities; it does not copy those rules into this package.
 
+Provider-facing `paperclip_finish` accepts an omitted or `null` continuation for
+`done`, `completed`, and `needs_review`. This supports gateways that require all
+declared tool properties to be present. Normalization removes only `null`;
+non-yielding tool calls still reject a continuation object, and `yielded` still
+requires a complete `response_wake` object.
+
 ## Quick start
 
 Native provider debug-trace correlation uses an incremental index owned by its transport. Pending
diff --git a/packages/paperclip-runner/src/contracts/completion-result.test.ts b/packages/paperclip-runner/src/contracts/completion-result.test.ts
index c23a7e8563..7fd02435ba 100644
--- a/packages/paperclip-runner/src/contracts/completion-result.test.ts
+++ b/packages/paperclip-runner/src/contracts/completion-result.test.ts
@@ -131,6 +131,19 @@ describe("provider-neutral completion result schema", () => {
     expect(validate(response)).toBe(false);
   });
 
+  it.each(["done", "needs_review", "completed"])("accepts explicit no-continuation at the provider boundary for %s", (disposition) => {
+    const providerValidate = new Ajv2020({ allErrors: true, strict: false })
+      .compile(PRP_COMPLETION_RESULT_PROVIDER_INPUT_SCHEMA);
+    const response = {
+      ...structuredClone(baseResult),
+      reportedWorkDisposition: disposition,
+      continuation: null,
+    };
+    expect(providerValidate(response)).toBe(true);
+    // The canonical output remains strict; normalization removes only null.
+    expect(validate(response)).toBe(false);
+  });
+
   it("exposes concrete completion fields while retaining response-wake validation", () => {
     // The live Codex code-mode renderer reduced a conditional-only root allOf
     // to `args: unknown`. Keep this tool object-shaped for provider discovery.
@@ -154,6 +167,7 @@ describe("provider-neutral completion result schema", () => {
     };
     expect(providerValidate(yielded)).toBe(true);
     expect(providerValidate({ ...yielded, continuation: undefined })).toBe(false);
+    expect(providerValidate({ ...yielded, continuation: null })).toBe(false);
     expect(providerValidate({ ...yielded, continuation: { kind: "response_wake" } })).toBe(false);
     expect(providerValidate({
       ...yielded, continuation: { ...yielded.continuation, kind: "same_agent" },
diff --git a/packages/paperclip-runner/src/contracts/completion-result.ts b/packages/paperclip-runner/src/contracts/completion-result.ts
index dec93a2f1d..6beca921f2 100644
--- a/packages/paperclip-runner/src/contracts/completion-result.ts
+++ b/packages/paperclip-runner/src/contracts/completion-result.ts
@@ -355,15 +355,23 @@ export const PRP_COMPLETION_RESULT_PROVIDER_INPUT_SCHEMA = {
     reportedWorkDisposition: { enum: ["done", "needs_review", "yielded", "completed"] },
     verification: providerVerificationCompatibilitySchema,
     attentionRequests: providerAttentionCompatibilitySchema,
-    continuation: responseWakeContinuationSchema,
+    // Responses-compatible gateways can require every declared property in
+    // tool calls. Give non-yielding results an explicit absence value instead
+    // of forcing callers to invent a response-wake continuation.
+    continuation: {
+      ...responseWakeContinuationSchema,
+      type: ["object", "null"],
+      description:
+        "Use null or omit this field for done, completed, or needs_review. Only yielded requires a response-wake object with kind, summary, and idempotencyKey.",
+    },
   },
   // Keep the provider-facing root a concrete object. Codex code-mode renders a
   // root allOf containing only an if/then constraint as `args: unknown`, hiding
   // every required field from the model. This equivalent direct conditional
   // preserves validation without obscuring the object-shaped tool signature.
   if: { properties: { reportedWorkDisposition: { const: "yielded" } }, required: ["reportedWorkDisposition"] },
-  then: { required: ["continuation"] },
-  else: { not: { required: ["continuation"] } },
+  then: { required: ["continuation"], properties: { continuation: { type: "object" } } },
+  else: { properties: { continuation: { type: "null" } } },
 } as const;
 
 export const PRP_BLOCK_RESULT_PROVIDER_INPUT_SCHEMA = {
diff --git a/packages/paperclip-runner/src/protocol/result-normalization.test.ts b/packages/paperclip-runner/src/protocol/result-normalization.test.ts
index fcce070e36..7128beae7d 100644
--- a/packages/paperclip-runner/src/protocol/result-normalization.test.ts
+++ b/packages/paperclip-runner/src/protocol/result-normalization.test.ts
@@ -6,6 +6,36 @@ import {
 import { validatePrpStructuredRunResult } from "./replay-contract.js";
 
 describe("normalizePrpResultSignals", () => {
+  it("normalizes an explicit null continuation without weakening completion rules", () => {
+    const providerResult = {
+      reportedWorkDisposition: "done",
+      summary: "Renamed the task to bobathan.",
+      completionClaim: {
+        contractRevision: "1",
+        objectiveSatisfied: true,
+        criteria: [{ criterionId: "objective", status: "satisfied", evidenceRefs: [] }],
+        remainingWork: [],
+      },
+      evidence: [],
+      verification: [],
+      continuation: null,
+    };
+    const validated = validatePrpStructuredRunResult(providerResult);
+    expect(validated).toMatchObject({ ok: true, result: { reportedWorkDisposition: "done" } });
+    if (validated.ok) expect(validated.result).not.toHaveProperty("continuation");
+    expect(providerResult).toHaveProperty("continuation", null);
+    expect(validatePrpStructuredRunResult({
+      ...providerResult,
+      reportedWorkDisposition: "yielded",
+    })).toMatchObject({ ok: false });
+    expect(normalizeLegacyPrpStructuredRunResult({
+      ...providerResult,
+      continuation: { kind: "response_wake", summary: "Contradictory wait.", idempotencyKey: "wait-1" },
+    })).toHaveProperty("continuation", {
+      kind: "response_wake", summary: "Contradictory wait.", idempotencyKey: "wait-1",
+    });
+  });
+
   it("accepts legacy completion aliases as one canonical PRP result", () => {
     const legacy = {
       schema: "paperclip.run_result.v1",
diff --git a/packages/paperclip-runner/src/protocol/result-normalization.ts b/packages/paperclip-runner/src/protocol/result-normalization.ts
index 47ab63c309..ae27d13bf1 100644
--- a/packages/paperclip-runner/src/protocol/result-normalization.ts
+++ b/packages/paperclip-runner/src/protocol/result-normalization.ts
@@ -119,6 +119,12 @@ export function normalizeLegacyPrpStructuredRunResult(value: unknown): unknown {
   const source = record(value);
   if (Object.keys(source).length === 0) return value;
   const normalized: Record<string, unknown> = { ...source };
+  // Some provider transports require all tool properties to be present. Null
+  // explicitly means no continuation; preserve every non-null value so strict
+  // canonical validation still rejects contradictory or malformed waits.
+  if (normalized.continuation === null) {
+    delete normalized.continuation;
+  }
   // Provider tool-callers do not consistently echo constant discriminator
   // fields even when they are present in the advertised JSON schema. The
   // transport already selected the semantic-result tool, so adding its