Keep benchmark runs independent of PR comments

Author: rgarcia

.github/workflows/benchmark-clawbench.yml

@@ -200,6 +200,7 @@ jobs:
     permissions:
       contents: read
       issues: write
+      pull-requests: write
     concurrency:
       group: benchmark-clawbench-${{ needs.resolve.outputs.pr_number || needs.resolve.outputs.head_sha }}
       cancel-in-progress: false
@@ -231,6 +232,7 @@ jobs:
     steps:
       - name: Mark the PR benchmark as running
         if: needs.resolve.outputs.pr_number != ''
+        continue-on-error: true
         uses: actions/github-script@f28e40c7f34bde8b3046d885e986cb6290c5673b # v7.1.0
         env:
           PR_NUMBER: ${{ needs.resolve.outputs.pr_number }}
@@ -434,6 +436,7 @@ jobs:
 
       - name: Update PR benchmark comment
         if: always() && needs.resolve.outputs.pr_number != ''
+        continue-on-error: true
         uses: actions/github-script@f28e40c7f34bde8b3046d885e986cb6290c5673b # v7.1.0
         env:
           PR_NUMBER: ${{ needs.resolve.outputs.pr_number }}

benchmarks/harbor/results.test.ts

@@ -369,6 +369,13 @@ describe("benchmark workflow hardening", () => {
     expect(workflow).not.toContain("baseSha = pull.base.sha");
     expect(workflow).toContain('HARBOR_VERSION: "0.21.0"');
     expect(workflow).toContain('CODEX_BENCHMARK_VERSION: "0.120.0"');
+    expect(workflow).toContain("issues: write\n      pull-requests: write");
+    expect(workflow).toMatch(
+      /- name: Mark the PR benchmark as running\n\s+if:.*\n\s+continue-on-error: true/,
+    );
+    expect(workflow).toMatch(
+      /- name: Update PR benchmark comment\n\s+if:.*\n\s+continue-on-error: true/,
+    );
     expect(workflow).toContain(
       'statuses=(--status "candidate=${CANDIDATE_STATUS:-1}")',
     );