@crawlcheck/sdk 1.15.1 → 1.15.2

package/dist/index.d.ts

@@ -1136,3 +1136,5 @@ export interface LangChainToolsModule {
  * Same three tools as agentTools(); each returns its result as a JSON string; they fail closed. */
 export declare function langchainTools(cc: CrawlCheck, lc: LangChainToolsModule, opts?: AgentToolOptions): unknown[];
 export default CrawlCheck;
+/** Headers guardedFetch still sends after a redirect leaves the origin the caller named. Everything else stays behind. */
+export declare const GUARDED_SAFE_HEADERS: string[];

package/dist/index.js

@@ -495,7 +495,8 @@ export class CrawlCheck {
         const action = opts.action ?? "read", max = opts.maxHops ?? 10, f = opts.fetch ? opts.fetch.bind(globalThis) : this.f;
         const init = Object.assign({}, opts.init || {}), method = String(init.method || "GET").toUpperCase();
         const hops = [], seen = new Map();
-        let cur = url;
+        let cur = url, origin0 = "";
+        try { origin0 = new URL(url).origin; } catch { origin0 = ""; }
         for (let n = 0; n <= max; n++) {
             let u;
             try {
@@ -519,7 +520,10 @@ export class CrawlCheck {
             hops.push(hop);
             if (!g.proceed)
                 return { proceed: false, response: null, url: cur, hops, why: [dom + ": " + (g.why.join("; ") || "refused")] };
-            const h = new Headers(init.headers || {});
+            // hdr1 (1.15.2): caller headers stay on the origin the caller named; elsewhere only the allowlist travels
+            const sameOrigin = u.origin === origin0;
+            const h = new Headers();
+            new Headers(init.headers || {}).forEach((v, k) => { if (sameOrigin || GUARDED_SAFE_HEADERS.includes(k.toLowerCase())) h.set(k, v); });
             if (g.header)
                 for (const [k, v] of Object.entries(g.header))
                     h.set(k, v);
@@ -791,3 +795,5 @@ export function langchainTools(cc, lc, opts = {}) {
     return agentTools(cc, opts).map((t) => lc.tool(async (input) => JSON.stringify(await t.execute(input ?? {})), { name: t.name, description: t.description, schema: t.parameters }));
 }
 export default CrawlCheck;
+/** Headers guardedFetch still sends after a redirect leaves the origin the caller named. */
+export const GUARDED_SAFE_HEADERS = ["accept", "accept-language", "accept-encoding", "user-agent", "cache-control", "pragma"];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@crawlcheck/sdk",
-  "version": "1.15.1",
+  "version": "1.15.2",
   "description": "Typed client for the CrawlCheck API (generated from its OpenAPI document) and the zero-dependency offline verifier for evidence bundles and remediation receipts.",
   "type": "module",
   "main": "./dist/index.js",

package/test/guarded-headers.test.js

@@ -0,0 +1,32 @@
+// hdr1: caller headers stay on the origin the caller named; after a redirect to another origin only the allowlist travels.
+import test from "node:test"; import assert from "node:assert/strict";
+import { CrawlCheck, GUARDED_SAFE_HEADERS } from "../dist/index.js";
+function client() { const cc = new CrawlCheck(); cc.guard = async () => ({ proceed: true, why: [], receipt: { decision: "allow" }, header: { "crawlcheck-receipt": "r" } }); return cc; }
+function recorder(map) { const seen = []; const f = async (u, init) => { seen.push({ url: String(u), headers: Object.fromEntries(new Headers(init.headers).entries()) }); const to = map[String(u)]; return to ? new Response(null, { status: 302, headers: { location: to } }) : new Response("ok", { status: 200 }); }; return { f, seen }; }
+const H = { authorization: "Bearer secret", cookie: "sid=1", "x-api-key": "k", accept: "text/html", "user-agent": "t/1" };
+test("another host: Authorization, Cookie and custom headers are dropped; the allowlist and the receipt stay", async () => {
+  const { f, seen } = recorder({ "https://a.example/x": "https://b.example/y" });
+  const r = await client().guardedFetch("https://a.example/x", { init: { headers: H }, fetch: f });
+  assert.equal(r.proceed, true); assert.equal(seen.length, 2);
+  assert.equal(seen[0].headers.authorization, "Bearer secret"); assert.equal(seen[0].headers.cookie, "sid=1");
+  for (const k of ["authorization", "cookie", "x-api-key"]) assert.equal(seen[1].headers[k], undefined, k + " crossed origins");
+  assert.equal(seen[1].headers.accept, "text/html"); assert.equal(seen[1].headers["user-agent"], "t/1"); assert.equal(seen[1].headers["crawlcheck-receipt"], "r");
+});
+test("same origin: every caller header is kept", async () => {
+  const { f, seen } = recorder({ "https://a.example/x": "https://a.example/y" });
+  await client().guardedFetch("https://a.example/x", { init: { headers: H }, fetch: f });
+  assert.equal(seen[1].headers.authorization, "Bearer secret"); assert.equal(seen[1].headers["x-api-key"], "k");
+});
+test("scheme downgrade and port change count as another origin", async () => {
+  for (const to of ["http://a.example/y", "https://a.example:8443/y"]) {
+    const { f, seen } = recorder({ "https://a.example/x": to });
+    await client().guardedFetch("https://a.example/x", { init: { headers: H }, fetch: f });
+    assert.equal(seen[1].headers.authorization, undefined, to); assert.equal(seen[1].headers.cookie, undefined, to);
+  }
+});
+test("back to the first origin: headers return", async () => {
+  const { f, seen } = recorder({ "https://a.example/x": "https://b.example/y", "https://b.example/y": "https://a.example/z" });
+  await client().guardedFetch("https://a.example/x", { init: { headers: H }, fetch: f });
+  assert.equal(seen[1].headers.authorization, undefined); assert.equal(seen[2].headers.authorization, "Bearer secret");
+});
+test("the allowlist holds no credential header", () => { for (const k of ["authorization", "cookie", "proxy-authorization"]) assert.equal(GUARDED_SAFE_HEADERS.includes(k), false); });