Runner: parser, start(), diff/approve/reject, tested with real runs

runner/package.json

@@ -17,10 +17,12 @@
     "prepare": "tsc",
     "typecheck": "tsc --noEmit",
     "test": "vitest run",
-    "fake": "tsx scripts/fake.ts"
+    "fake": "tsx scripts/fake.ts",
+    "try": "tsx scripts/try.ts"
   },
   "dependencies": {
-    "@codehands/contracts": "file:../../codehands-contracts"
+    "@codehands/contracts": "file:../../codehands-contracts",
+    "execa": "^10.2.0"
   },
   "devDependencies": {
     "@types/node": "^26.6.5",

runner/scripts/try.ts

@@ -0,0 +1,43 @@
+// pnpm try [repoPath] [prompt] [--approve | --reject]: one real run through the runner, one event per line.
+// Without a flag it asks approve / reject / keep at the end (or keeps, when not in a terminal).
+// Costs real money (capped by budgetUsd below). Run scripts/reset-fixture.sh afterwards.
+import { randomUUID } from "node:crypto";
+import { homedir } from "node:os";
+import { join } from "node:path";
+import { createInterface } from "node:readline/promises";
+import { createRunner } from "../src/index.js";
+
+const flags = process.argv.slice(2).filter((a) => a.startsWith("--"));
+const [repoPath = join(homedir(), "codehands-fixtures", "date-bug"), prompt = "Fix the failing test. Run npm test to check."] =
+  process.argv.slice(2).filter((a) => !a.startsWith("--"));
+
+const runner = createRunner();
+const sessionId = randomUUID();
+let finished = false;
+for await (const e of runner.start({ sessionId, repoPath, prompt, agent: "claude-code", model: "sonnet", budgetUsd: 0.5 })) {
+  const p = e.payload;
+  const detail = p.reason ?? p.command ?? p.path ?? (typeof p.text === "string" ? p.text.slice(0, 80) : "") ?? "";
+  const extra = e.type === "test.result" ? ` passed=${p.passed}` : e.type === "task.finished" ? ` $${p.total_cost_usd}` : "";
+  console.log(`${String(e.seq).padStart(3)}  ${e.type.padEnd(17)} ${String(detail).replace(/\s+/g, " ")}${extra}`);
+  if (e.type === "task.finished") finished = true;
+}
+if (!finished) process.exit(1);
+
+console.log(`\n${(await runner.diff(sessionId)).stat}\n`);
+
+let choice = flags.includes("--approve") ? "a" : flags.includes("--reject") ? "r" : "k";
+if (!flags.length && process.stdin.isTTY) {
+  const rl = createInterface({ input: process.stdin, output: process.stdout });
+  choice = (await rl.question("approve, reject or keep? [a/r/k] ")).trim().toLowerCase()[0] ?? "k";
+  rl.close();
+}
+
+if (choice === "a") {
+  const { branch, commit } = await runner.approve(sessionId, prompt);
+  console.log(`approved: ${branch} @ ${commit.slice(0, 7)}  (git log ${branch} --oneline -1)`);
+} else if (choice === "r") {
+  await runner.reject(sessionId);
+  console.log("rejected: worktree and branch deleted");
+} else {
+  console.log("kept: the worktree is still there, see `git worktree list` in the repo");
+}