fix: skip step-checker in container jobs where _diag is not accessible

The step-checker was causing warnings in container jobs because the _diag
directory exists on the host but is not mounted into containers.

This change adds container detection via /proc/1/cgroup and skips the
step-checker gracefully when running inside a container, allowing sticky
disk commits to proceed normally.

Affects container jobs only - regular jobs continue to work as before.

Author: adityamaru

dist/post/index.js

@@ -23,6 +23,29 @@ export async function checkPreviousStepFailures(
   runnerBasePath?: string,
 ): Promise<StepFailureCheck> {
   try {
+    // Check if we're running inside a container.
+    // In container jobs, _diag is not mounted and not accessible.
+    const isContainer = await (async () => {
+      try {
+        const cgroup = await fs.readFile("/proc/1/cgroup", "utf-8");
+        return cgroup.includes("docker") || cgroup.includes("containerd");
+      } catch {
+        return false;
+      }
+    })();
+
+    if (isContainer) {
+      core.debug(
+        "Running inside container - _diag directory not accessible, skipping step failure check",
+      );
+      return {
+        hasFailures: false,
+        failedCount: 0,
+        error:
+          "Step failure checking skipped: running inside container where _diag is not accessible",
+      };
+    }
+
     // If no base path provided, try to detect the runner root
     if (!runnerBasePath) {
       // In GitHub Actions, we're typically in /home/runner/_work/{repo}/{repo}