[rig-claude] Improve Claude dynamic-workflow compatibility for rig

skills/rig/samples/320-budget-aware-crawler.md

@@ -0,0 +1,64 @@
+# 320 - Budget-Aware Crawler (log, budget, until)
+
+Demonstrates `log`, `budget`, and `until` — the three Claude dynamic-workflow
+primitives most likely to be ported without an obvious rig example. Maps
+directly to Claude's `log(message)`, `budget.remaining()`, and open-ended
+`while` convergence loops.
+
+See [claude-workflow-conversion.md](../references/claude-workflow-conversion.md)
+for the full primitive mapping.
+
+```rig
+import { s, workflow } from "rig";
+
+// Workflow role: crawl a starting URL, follow links, and summarise pages while
+// the agent budget allows, then return a structured index.
+const crawler = workflow({
+  meta: {
+    name: "crawler",
+    description: "Budget-aware web crawler with convergence loop",
+    phases: ["Seed", "Crawl"],
+    whenToUse: "Summarising an unknown number of pages when agent budget may be limited.",
+  },
+  input: s.object({ seedUrl: s.url }),
+  body: async ({ call, input, log, budget, until, phase }) => {
+    phase("Seed");
+    log(`Starting crawl from ${input.seedUrl}`);
+
+    const visited = new Set<string>();
+    const queue: string[] = [input.seedUrl];
+
+    phase("Crawl");
+    const pages: { url: string; summary: string }[] = [];
+
+    await until({ max: 20, noProgressRounds: 3 }, async (_, round) => {
+      const url = queue.shift();
+      if (!url || visited.has(url)) return { state: pages, done: queue.length === 0, progressKey: `${visited.size}` };
+
+      log(`Round ${round}: crawling ${url} (budget remaining: ${budget.remaining()})`);
+      visited.add(url);
+
+      const result = await call.json(
+        `Fetch and summarise the page at ${url}. Return a one-sentence summary and up to 3 new links found on the page.`,
+        s.object({ summary: s.string, links: s.array(s.url) }),
+        { label: url },
+      );
+
+      if (result) {
+        pages.push({ url, summary: result.summary });
+        for (const link of result.links) {
+          if (!visited.has(link)) queue.push(link);
+        }
+      }
+
+      const done = queue.length === 0 || budget.remaining() < 3;
+      if (done) log(`Crawl complete: ${pages.length} pages, ${queue.length} links remaining`);
+      return { state: pages, done, progressKey: `${visited.size}` };
+    });
+
+    return pages;
+  },
+});
+
+export default crawler;
+```