feat: add path validation options to restore action

__tests__/e2e/save-poisoned-cache.mjs

@@ -0,0 +1,45 @@
+// @ts-check
+/**
+ * save-poisoned-cache.mjs
+ *
+ * Helper script used by the path-validation E2E workflow to upload a cache
+ * archive that contains entries outside the declared `path` inputs. This
+ * simulates a poisoned cache that would have been produced by a build job
+ * that had write access to the workspace's parent directory (the canonical
+ * cache-poisoning scenario being defended against).
+ *
+ * Usage:
+ *   node save-poisoned-cache.mjs <cache-key> <declared-path> [extra-path ...]
+ *
+ * The script invokes `@actions/cache.saveCache()` with the declared path(s)
+ * AND extra paths that escape the workspace. The toolkit's saveCache packs
+ * everything into the archive, so the resulting cache entry will contain
+ * "escape" entries that resolve outside the declared `path` when the action's
+ * `restore` step later extracts it (because the restore step only declares the
+ * legitimate `path`).
+ *
+ * Important: this script is NOT shipped to users. It is purely a test fixture
+ * generator used by the E2E workflow to validate that the action's client-side
+ * validation correctly rejects (or warns about) such caches.
+ */
+
+import * as cache from '@actions/cache';
+
+const [, , key, ...paths] = process.argv;
+
+if (!key || paths.length === 0) {
+    console.error(
+        'Usage: node save-poisoned-cache.mjs <cache-key> <path> [extra-path ...]'
+    );
+    process.exit(2);
+}
+
+console.log(`Saving poisoned cache with key="${key}" paths=${JSON.stringify(paths)}`);
+
+try {
+    const cacheId = await cache.saveCache(paths, key);
+    console.log(`Saved poisoned cache (cacheId=${cacheId})`);
+} catch (err) {
+    console.error(`Failed to save poisoned cache: ${err?.message ?? err}`);
+    process.exit(1);
+}