diff --git a/.github/workflows/openclaw-release-telegram-qa.yml b/.github/workflows/openclaw-release-telegram-qa.yml
index fc18fe63c174..e5977f117283 100644
--- a/.github/workflows/openclaw-release-telegram-qa.yml
+++ b/.github/workflows/openclaw-release-telegram-qa.yml
@@ -1498,13 +1498,11 @@ jobs:
             local config_path="${1:?}"
             local config_id diagnostics_dir diagnostics_real proof_bytes proof_path proof_tmp
             umask 077
-            config_id="$(printf '%s' "$config_path" | sha256sum | awk '{print $1}')"
             diagnostics_dir="${EVIDENCE_ROOT}/trusted-runtime-diagnostics"
             install -d -o "$RUNNER_UID" -g "$RUNNER_GID" -m 0700 "$diagnostics_dir"
             [[ ! -L "$diagnostics_dir" ]]
             diagnostics_real="$(realpath -e "$diagnostics_dir")"
             [[ "$diagnostics_real" == "$diagnostics_dir" ]]
-            proof_path="${diagnostics_dir}/gateway-model-config-${config_id}.json"
             proof_tmp="${RUNTIME_ROOT}/gateway-model-config-${BASHPID}.json"
 
             # Capture only bounded routing identifiers before the QA suite removes its temp config.
@@ -1552,6 +1550,10 @@ jobs:
             ' "$config_path" >"$proof_tmp"
             proof_bytes="$(stat -c '%s' "$proof_tmp")"
             ((proof_bytes > 0 && proof_bytes <= 65536))
+            # Content addressing deduplicates credential-specific temp configs
+            # that project to the same bounded routing proof.
+            config_id="$(sha256sum "$proof_tmp" | awk '{print $1}')"
+            proof_path="${diagnostics_dir}/gateway-model-config-${config_id}.json"
             chown "$RUNNER_UID:$RUNNER_GID" "$proof_tmp"
             chmod 0600 "$proof_tmp"
             mv -f "$proof_tmp" "$proof_path"
diff --git a/AGENTS.md b/AGENTS.md
index ec005211e67a..08109296d877 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -215,6 +215,7 @@ Skills own workflows; root owns hard policy and routing.
 - Trusted-workflow release-branch CI: pass `target_ref` + `release_candidate_ref`; never `release_gate` (requires workflow head == target).
 - Agent PR landing to `main`: use only the repo-native `scripts/pr` wrapper: run `scripts/pr review-init <PR>`, follow its emitted checkout/guard guidance, initialize and complete review artifacts with `scripts/pr review-artifacts-init <PR>`, validate them with `scripts/pr review-validate-artifacts <PR>`, then run `OPENCLAW_TESTBOX=1 scripts/pr prepare-run <PR>` and `scripts/pr merge-run <PR>`. The Testbox flag is mandatory for agents so prepare verifies hosted CI/Testbox on the current head or reuses a patch-identical pre-rebase run green within 24 hours instead of running full gates locally. `prepare-run` fails fast; invoke only after exact-head CI is complete and green. For owner-approved reviewed fork code without hosted Testbox, use `OPENCLAW_PR_GATES_REMOTE=testbox` instead. Do not rebase only because `main` advanced; merge drift is advisory unless strict drift is explicitly enabled, while GitHub still blocks conflicts. Do not idle on `auto-response` or `check-docs`.
 - Non-main PRs: do not run `scripts/pr prepare-run` or `merge-run`; they diff against `main`. Use review artifacts, exact base-head CI, revalidate `headRefOid`, then `gh pr merge --match-head-commit <verified-sha>`.
+- Merge guard shells: start `set -euo pipefail`; a failed `[[ ... ]]` alone does not stop a later merge command.
 - After `scripts/pr merge-run` removes its worktree, `cd` to a persistent repo before follow-up commands.
 - `scripts/pr` review JSON: land-ready recommendation `READY FOR /prepare-pr`, `issueValidation.status=valid`; never `APPROVE`.
 
diff --git a/test/scripts/openclaw-release-telegram-qa-workflow.test.ts b/test/scripts/openclaw-release-telegram-qa-workflow.test.ts
index 9a685f2ad874..bf5c6b0a2395 100644
--- a/test/scripts/openclaw-release-telegram-qa-workflow.test.ts
+++ b/test/scripts/openclaw-release-telegram-qa-workflow.test.ts
@@ -792,6 +792,7 @@ describe("release Telegram QA workflow", () => {
       const runtimeRoot = join(workdir, "runtime");
       const evidenceRoot = join(workdir, "evidence");
       const configPath = join(workdir, "openclaw.json");
+      const duplicateConfigPath = join(workdir, "openclaw-duplicate.json");
       mkdirSync(runtimeRoot);
       mkdirSync(evidenceRoot);
       writeFileSync(
@@ -815,17 +816,22 @@ describe("release Telegram QA workflow", () => {
           },
         }),
       );
+      writeFileSync(
+        duplicateConfigPath,
+        readFileSync(configPath, "utf8").replace("not-exported", "different-ignored-value"),
+      );
       const result = spawnSync(
         "bash",
         [
           "-c",
-          `set -euo pipefail\n${captureSource}\ncapture_live_model_config "$CONFIG_PATH"\nfind "$EVIDENCE_ROOT/trusted-runtime-diagnostics" -type f -name 'gateway-model-config-*.json' -print`,
+          `set -euo pipefail\n${captureSource}\ncapture_live_model_config "$CONFIG_PATH"\ncapture_live_model_config "$DUPLICATE_CONFIG_PATH"\nfind "$EVIDENCE_ROOT/trusted-runtime-diagnostics" -type f -name 'gateway-model-config-*.json' -print`,
         ],
         {
           encoding: "utf8",
           env: {
             ...process.env,
             CONFIG_PATH: configPath,
+            DUPLICATE_CONFIG_PATH: duplicateConfigPath,
             EVIDENCE_ROOT: evidenceRoot,
             RUNTIME_ROOT: runtimeRoot,
             RUNNER_GID: String(process.getgid?.() ?? 0),
@@ -834,7 +840,9 @@ describe("release Telegram QA workflow", () => {
         },
       );
       expect(result.status, result.stderr).toBe(0);
-      const proofPath = result.stdout.trim();
+      const proofPaths = result.stdout.trim().split("\n");
+      expect(proofPaths).toHaveLength(1);
+      const proofPath = proofPaths[0]!;
       const proofText = readFileSync(proofPath, "utf8");
       expect(JSON.parse(proofText)).toEqual({
         agentDefaultModel: {