fix(story-mode): "Done, with a catch" only for a problem that is still open

A sweep of 1,314 real transcripts (3,863 pages) titled 330 of them
"Done, with a catch"; a sample showed about half were history or
negations: "I found and fixed a model bug", "Bug fixed: …", "found no
broken links", "Not a problem", "the caveat is gone", "was never
broken", plus names ("bug-fixing", "cron-broken-network",
"#limitations", "the memory warning").

A problem sentence now counts only while it is open: not described as
fixed, gone or the past cause (unless it says "not fixed"), not a
negation, and a bare "warning" no longer counts. "Left out" and "left
open" are open work (a note), unless Claude chose it ("left out on
purpose", "rather than guessed"). 244 pages keep the catch; samples
read as real open problems in about four of five.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

test/story-mode.test.ts

@@ -13,6 +13,7 @@ import {
   exactDuration,
   firstSentence,
   heredocWrites,
+  keyKindOf,
   parseTestCounts,
   reportsProblems,
   planFromText,
@@ -1004,6 +1005,43 @@ describe('buildStoryPage', () => {
     expect(run('git config user.name; ./deploy.sh', 'Exit code 1\nme')).toBe('Claude tries a command, and it fails.');
   });
 
+  it('a problem in the last words is an open one: not history, not a negation, not a name', () => {
+    // Open: a catch.
+    for (const open of [
+      "Prompts sent through the API didn't submit (separate bug, not fixed).",
+      'The main problem is RAM: the box has 30 GB.',
+      "Opening the page: it won't work if you double-click index.html.",
+      'Limitation: the page does not follow the OS theme until a reload.',
+      'The bug is still on master.',
+    ])
+      expect(keyKindOf(open), open).toBe('problem');
+    // History, a negation or a name: no catch.
+    for (const closed of [
+      'I found and fixed a model bug.',
+      'Bug fixed: the server quietly rewrote a leading slash.',
+      "The bridge's model-name problem is fixed by infra-guard in commit 19081f6.",
+      'This also fixes an old bug: an HTTP error used to show the wrong toast.',
+      'The old sockets were the problem: they lacked that option.',
+      'The "don\'t write confirmed yet" caveat is gone.',
+      "The build's link check found no broken links.",
+      'Not a problem, just not the one-shot I described.',
+      "I couldn't find a real bug in any of these four.",
+      'The shutdown path was never broken.',
+      "It isn't a key problem or a wrong username.",
+      'SWE-bench is real bug-fixing in large repositories.',
+      'Requirements are listed at https://example.com/docs#limitations today.',
+      'A bug fix this round turned up.',
+      'The memory warning now fires at 90 percent.',
+    ])
+      expect(['problem'], closed).not.toContain(keyKindOf(closed));
+    // "Left out" is open work, unless Claude chose it.
+    expect(keyKindOf('The README update is left out for now.')).toBe('pending');
+    expect(
+      keyKindOf('Left out on purpose: the search in the working tree belongs to another session.')
+    ).toBeUndefined();
+    expect(keyKindOf('Anything unknown is left out rather than guessed.')).toBeUndefined();
+  });
+
   it('a stop names the step Claude was on and what its to-do list still had open', () => {
     const stopped = (lines: unknown[]) =>
       pageOf([...lines, prompt('[Request interrupted by user]')]).panels.find((p) => p.kind === 'complete')!.caption