fix(autoplan): open Critical and High findings reach the gate as apply items (GBRA-223 claim 3)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

test/autoplan-answer-invalidation.test.ts

@@ -13,12 +13,12 @@ import { spawnSync } from 'node:child_process';
 import * as fs from 'node:fs';
 import * as os from 'node:os';
 import * as path from 'node:path';
-import { attemptsOf, autoplan, journal, readRun, resultOf, review, ROOT, runPhase, SMALL_PLAN, startSmallRun } from './helpers/autoplan-run-fixture';
+import { attemptsOf, autoplan, journal, readRun, recCarrying, resultOf, review, ROOT, runPhase, SMALL_PLAN, startSmallRun } from './helpers/autoplan-run-fixture';
 
 const TMP = fs.mkdtempSync(path.join(os.tmpdir(), 'autoplan-answer-'));
 afterAll(() => fs.rmSync(TMP, { recursive: true, force: true }));
 
-/** A gated run with a disagreement in ceo (taste item d1) and the fixture's user challenges (uc1 ceo, uc2 eng). */
+/** A gated run with a disagreement in ceo (taste item d1), the fixture's user challenges (uc1 ceo, uc2 eng) and its open Critical/High findings (d2 ceo, d3 and d4 eng). */
 function gatedRun(name: string, extra: string[] = []): string {
   const out = startSmallRun(path.join(TMP, name), extra);
   const disagree = JSON.stringify([
@@ -60,11 +60,12 @@ function incorporate(out: string, ids: string[], edit: string) {
 }
 
 describe('answer', () => {
-  test('the gate lists the disagreement as an auto item and the user challenges as approval items', () => {
+  test('the gate lists the disagreement and the open Critical/High findings as auto items and the user challenges as approval items', () => {
     const out = gatedRun('shape');
     const gate = JSON.parse(fs.readFileSync(path.join(out, 'gate.json'), 'utf8'));
     expect(gate.gate_rev).toBe(1);
-    expect(gate.items.map((i: any) => [i.id, i.kind, i.phase ?? null])).toEqual([['d1', 'auto', 'ceo'], ['uc1', 'approval', 'ceo'], ['uc2', 'approval', 'eng'], ['p1', 'approval', null]]);
+    expect(gate.items.map((i: any) => [i.id, i.kind, i.phase ?? null])).toEqual([['d1', 'auto', 'ceo'], ['uc1', 'approval', 'ceo'], ['d2', 'auto', 'ceo'], ['uc2', 'approval', 'eng'], ['d3', 'auto', 'eng'], ['d4', 'auto', 'eng'], ['p1', 'approval', null]]);
+    expect(gate.items.filter((i: any) => i.finding).map((i: any) => i.id)).toEqual(['d2', 'd3', 'd4']);
     expect(gate.items[0].why).toMatch(/voices disagree: severity: native High, outside Low/);
   }, 30_000);
 
@@ -95,17 +96,18 @@ describe('answer', () => {
     expect(rows.filter((r: any) => r.status === 'pending').map((r: any) => r.label)).toEqual(['UC1', 'UC2', 'P1']);
   }, 30_000);
 
-  test('`all` finalizes without another paid phase: status complete, decisions approved, no phase reopened', () => {
+  test('`all` carrying the open findings finalizes without another paid phase: status complete, decisions answered, no phase reopened', () => {
     const out = gatedRun('approve');
     const counter = readRun(out).attempt_counter;
-    const r = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '1', '--reply', 'all']);
+    const reply = recCarrying(out);
+    const r = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '1', '--reply', reply]);
     expect(r.status, r.stderr).toBe(0);
     expect(r.stdout).toMatch(/^APPROVED: gate_rev=1 as-is$/m);
     expect(r.stdout).toContain('GSTACK_RESULT: skill=autoplan status=complete run=' + out);
     const run = readRun(out);
     expect(run.status).toBe('complete');
     expect(run.attempt_counter).toBe(counter);
-    expect(decisions(out).every((d: any) => d.status === 'approved' && d.reply === 'all')).toBe(true);
+    expect(decisions(out).every((d: any) => d.status === (/^D[234]$/.test(d.label) ? 'overridden' : 'approved') && d.reply === reply)).toBe(true);
     expect(journal(out).filter(j => j.event === 'reopened')).toEqual([]);
     const again = autoplan(['next', '--out', out]);
     expect(again.stdout).toContain('status=complete');
@@ -114,7 +116,7 @@ describe('answer', () => {
 
   test('a non-recommended choice on a ceo item is recorded without reopening; revise reopens the changed phases, archives their bound files, and the next gate is gate_rev 2', () => {
     const out = gatedRun('reopen', ['--no-carry-forward']);
-    const r = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '1', '--reply', 'all except d1b', '--json']);
+    const r = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '1', '--reply', recCarrying(out, 'all', ['d1b']), '--json']);
     expect(r.status, r.stderr).toBe(0);
     const runId = readRun(out).run;
     expect(JSON.parse(r.stdout)).toMatchObject({ status: 'amend_pending', outcome: 'amend_pending', reopened: [], amendments: [`${runId}-d1`] });
@@ -145,12 +147,12 @@ describe('answer', () => {
     expect(eng.outside.stdout).toContain('status=gate_pending');
     expect(readRun(out).gate_rev).toBe(2);
     const rows = decisions(out);
-    expect(rows.filter((x: any) => x.gate_rev === 1)).toHaveLength(4);
+    expect(rows.filter((x: any) => x.gate_rev === 1)).toHaveLength(7);
     expect(rows.filter((x: any) => x.gate_rev === 2 && x.status === 'pending').length).toBeGreaterThan(0);
     const staleAgain = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '1', '--reply', 'all']);
     expect(staleAgain.status).toBe(3);
     // closing gate_rev 2 exports a run that validates: rev-2 rows get their own ids beside rev 1's
-    const approve = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '2', '--reply', 'all']);
+    const approve = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '2', '--reply', recCarrying(out)]);
     expect(approve.stdout).toContain('GSTACK_RESULT: skill=autoplan status=complete run=' + out);
     expect(autoplan(['export', '--out', out]).status).toBe(0);
     const ids = decisions(out).map((x: any) => x.id);
@@ -167,7 +169,7 @@ describe('answer', () => {
 
   test('an eng-only override reopens nothing at answer; revise reruns every phase that reviewed the old plan; p1d reopens every phase; p1c keeps the gate; p1e ends incomplete', () => {
     const out = gatedRun('eng-only', ['--max-rounds', '3', '--no-carry-forward']);
-    const engOnly = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '1', '--reply', 'all except uc2b', '--json']);
+    const engOnly = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '1', '--reply', recCarrying(out, 'all', ['uc2b']), '--json']);
     const runId = readRun(out).run;
     expect(JSON.parse(engOnly.stdout)).toMatchObject({ outcome: 'amend_pending', reopened: [], amendments: [`${runId}-uc2`] });
     expect(readRun(out).phases.ceo.closed_at).toBeTruthy();
@@ -178,13 +180,13 @@ describe('answer', () => {
     expect(runPhase(out, 'ceo').outside.stdout).toMatch(/PHASE_CLOSED: ceo/);
     const eng = runPhase(out, 'eng');
     expect(eng.outside.stdout).toContain('status=gate_pending');
-    const interrogate = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '2', '--reply', 'all except p1c', '--json']);
+    const interrogate = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '2', '--reply', recCarrying(out, 'all', ['p1c']), '--json']);
     expect(JSON.parse(interrogate.stdout)).toMatchObject({ status: 'gate_pending', outcome: 'gate_pending' });
-    const revise = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '2', '--reply', 'all except p1d', '--json']);
+    const revise = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '2', '--reply', recCarrying(out, 'all', ['p1d']), '--json']);
     expect(JSON.parse(revise.stdout)).toMatchObject({ outcome: 'reopened', reopened: ['ceo', 'eng'] });
     runPhase(out, 'ceo');
     runPhase(out, 'eng');
-    const reject = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '3', '--reply', 'all except p1e']);
+    const reject = autoplan(['answer', '--by', 'human', '--out', out, '--gate-rev', '3', '--reply', recCarrying(out, 'all', ['p1e'])]);
     expect(reject.status).toBe(0);
     expect(reject.stdout).toContain('GSTACK_RESULT: skill=autoplan status=incomplete run=' + out);
     expect(readRun(out).deviations).toContain('plan rejected by the owner at gate_rev 3');

test/autoplan-capy-e2e.test.ts

@@ -12,7 +12,7 @@ import { afterAll, describe, expect, test } from 'bun:test';
 import * as fs from 'node:fs';
 import * as path from 'node:path';
 import { RESULT_CODES } from '../lib/result-codes';
-import { ROOT, SMALL_PLAN, boundOf, resultOf } from './helpers/autoplan-run-fixture';
+import { ROOT, SMALL_PLAN, boundOf, recCarrying, resultOf } from './helpers/autoplan-run-fixture';
 import { capyRunner, phaseVia, readJsonl, readRunJson, writePlan, type CapyRunner } from './helpers/autoplan-capy';
 
 const r = capyRunner();
@@ -55,11 +55,11 @@ describe('unattended runs on the Capy host', () => {
       phaseVia(r, s.out, 'ceo');
       expect(phaseVia(r, s.out, 'eng').outside.stdout).toContain('status=gate_pending');
       expectReviewedHashes(s.out);
-      expect(ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', String(rev), '--reply', 'all except p1d'])).stdout).toMatch(/^REOPENED: ceo,eng \(Eng last\) reason=revise$/m);
+      expect(ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', String(rev), '--reply', recCarrying(s.out, 'all', ['p1d'])])).stdout).toMatch(/^REOPENED: ceo,eng \(Eng last\) reason=revise$/m);
     }
     phaseVia(r, s.out, 'ceo');
     phaseVia(r, s.out, 'eng');
-    expect(ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '3', '--reply', 'all'])).stdout).toContain('status=complete');
+    expect(ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '3', '--reply', recCarrying(s.out)])).stdout).toContain('status=complete');
     ok(r.ap(['export', '--out', s.out]));
     const v = ok(r.validate(s.out));
     expect(v.stdout).toMatch(/^ARTIFACT_VALID: run=\S+ status=complete/m);
@@ -83,7 +83,7 @@ describe('unattended runs on the Capy host', () => {
     const gated = ok(r.ap(['submit', '--out', s.out, '--phase', 'eng', '--voice', 'outside', '--unavailable', 'no cross-family subagent on this host']));
     expect(gated.stdout).toMatch(/^PHASE_CLOSED: eng .*coverage=native_only/m);
     expect(gated.stdout).toContain('status=gate_pending');
-    expect(ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '1', '--reply', 'all'])).stdout).toContain('status=complete');
+    expect(ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '1', '--reply', recCarrying(s.out)])).stdout).toContain('status=complete');
     ok(r.ap(['export', '--out', s.out]));
     expect(ok(r.validate(s.out)).stdout).toMatch(/^ARTIFACT_VALID: run=\S+ status=complete attested=1$/m);
     expect(fs.readFileSync(path.join(s.out, 'review-record.md'), 'utf8')).toMatch(/attested/);
@@ -94,7 +94,7 @@ describe('unattended runs on the Capy host', () => {
     const s = start();
     phaseVia(r, s.out, 'ceo');
     phaseVia(r, s.out, 'eng');
-    ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '1', '--reply', 'all except p1d']));
+    ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '1', '--reply', recCarrying(s.out, 'all', ['p1d'])]));
     phaseVia(r, s.out, 'ceo');
     phaseVia(r, s.out, 'eng');
     const file = path.join(s.out, 'decisions.jsonl');
@@ -102,7 +102,7 @@ describe('unattended runs on the Capy host', () => {
     const manifest = readRunJson(s.out);
     delete manifest.gstack_version;
     fs.writeFileSync(path.join(s.out, 'run.json'), JSON.stringify(manifest, null, 2) + '\n');
-    const res = ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '2', '--reply', 'all']));
+    const res = ok(r.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '2', '--reply', recCarrying(s.out)]));
     expect(res.stdout).toMatch(/^GSTACK_VERSION_CHANGED: run unknown -> /m);
     expect(res.stdout).toContain('status=complete');
     ok(r.ap(['export', '--out', s.out]));
@@ -150,7 +150,7 @@ describe('conformance: one run through every mechanism', () => {
     const accept = JSON.parse(fs.readFileSync(path.join(s.out, 'gate.json'), 'utf8')).items.find((i: any) => i.title.startsWith('Gate the backoff')).id;
     // partial, then full
     expect(ok(c.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '1', '--reply', `${accept}b`])).stdout).toMatch(/^GATE_PENDING: p1/m);
-    const full = ok(c.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '1', '--reply', 'all']));
+    const full = ok(c.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '1', '--reply', recCarrying(s.out)]));
     expect(full.stdout).toMatch(new RegExp(`^APPLY: ${run}-${accept} `, 'm'));
     const pendingNext = c.ap(['next', '--out', s.out]).stderr;
     expect(pendingNext).toMatch(/\(PLAN_AMEND_PENDING\)$/m);
@@ -169,10 +169,10 @@ describe('conformance: one run through every mechanism', () => {
     for (const voice of ['native', 'outside']) ok(c.ap(['submit', '--out', s.out, '--phase', 'ceo', '--voice', voice, '--result', resultOf(s.out, 'ceo', voice as 'native' | 'outside'), '--model', voice === 'native' ? 'anthropic/claude-opus-5-5' : 'openai/gpt-6.1-sol']));
     phaseVia(c, s.out, 'eng');
     expect(readRunJson(s.out).gate_rev).toBe(2);
-    ok(c.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '2', '--reply', 'all except p1d']));
+    ok(c.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '2', '--reply', recCarrying(s.out, 'all', ['p1d'])]));
     phaseVia(c, s.out, 'ceo');
     phaseVia(c, s.out, 'eng');
-    expect(ok(c.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '3', '--reply', 'all'])).stdout).toContain('status=complete');
+    expect(ok(c.ap(['answer', '--by', 'human', '--out', s.out, '--gate-rev', '3', '--reply', recCarrying(s.out)])).stdout).toContain('status=complete');
     ok(c.ap(['export', '--out', s.out]));
     expect(ok(c.validate(s.out)).stdout).toMatch(/^ARTIFACT_VALID: run=\S+ status=complete/m);
     const final = readRunJson(s.out);

test/autoplan-gate-cost.test.ts

@@ -188,7 +188,7 @@ describe('E16 end to end: every gate-producing command prints the cost, and the
 
 
   test('the closing submit prints GATE_SPEND/GATE_REC_COST (text) or gate_cost (--json); the owner brief carries the Cost so far sentence and render --captain a ### Cost section above the relaying trailer', () => {
-    const out = gateRun('plain', {}, { challenge: false, json: true });
+    const out = gateRun('plain', { severity: 'Medium' }, { challenge: false, json: true });
     const brief = spawnSync(process.execPath, [path.join(ROOT, 'bin', 'gstack-owner-brief'), 'render', '--out', out], { cwd: ROOT, env: env(), encoding: 'utf8', timeout: 30_000 });
     expect(brief.stdout).toMatch(/Cost so far: spent \d+\.\d\d, worst case \d+\.\d\d \(/);
     const captain = spawnSync(process.execPath, [path.join(ROOT, 'bin', 'gstack-owner-brief'), 'render', '--captain', '--out', out, '--allow-incomplete'], { cwd: ROOT, env: env(), encoding: 'utf8', timeout: 30_000 });
@@ -203,7 +203,7 @@ describe('E16 end to end: every gate-producing command prints the cost, and the
   });
 
   test('run at a pending gate prints the gate lines once: text mode carries GATE_SPEND/GATE_REC_COST, --json exactly one payload with gate_cost', () => {
-    const out = gateRun('run-gate', {}, { challenge: false });
+    const out = gateRun('run-gate', { severity: 'Medium' }, { challenge: false });
     const models = ['--native-model', 'anthropic/claude-opus-4-7', '--outside-model', 'openai/gpt-6-astra'];
     const text = autoplan(['run', '--out', out, ...models]);
     expect(text.status).toBe(0);

test/autoplan-gate-judgment.test.ts

@@ -17,7 +17,7 @@ import { afterAll, describe, expect, test } from 'bun:test';
 import { spawnSync } from 'node:child_process';
 import * as fs from 'node:fs';
 import * as path from 'node:path';
-import { ROOT, SMALL_PLAN, boundOf, resultOf } from './helpers/autoplan-run-fixture';
+import { ROOT, SMALL_PLAN, boundOf, recCarrying, resultOf } from './helpers/autoplan-run-fixture';
 import { capyRunner, phaseVia, readJsonl, readRunJson, writePlan } from './helpers/autoplan-capy';
 
 const r = capyRunner();
@@ -119,7 +119,7 @@ describe('accepted changes are amendments', () => {
     expect(held.status, held.stderr).toBe(0);
     expect(held.stdout).toMatch(/^GATE_PENDING: uc2, p1 unanswered \(reply with uc2a\|uc2b p1a\|p1b\|p1c\|p1d\|p1e or rec\)$/m);
     expect(held.stdout).not.toMatch(/^APPLY:/m);
-    const res = answer(g.out, 1, 'rec');
+    const res = answer(g.out, 1, recCarrying(g.out, 'rec'));
     expect(res.status, res.stderr).toBe(0);
     expect(res.stdout).toMatch(new RegExp(`^APPLY: ${esc(g.run)}-uc1 Ship the backoff behind a flag first .* -> accept the change both reviewers recommend.*; edit the plan, then revise$`, 'm'));
     expect(res.stdout).not.toMatch(/^REOPENED:/m);
@@ -165,7 +165,7 @@ describe('accepted changes are amendments', () => {
     expect(asIs.stdout).not.toMatch(/^APPLY:/m);
     expect(decisions(g.out).find(d => d.label === 'UC1')).toMatchObject({ chosen: 'a', status: 'approved' });
     const w = gated(AGREE, {});
-    expect(answer(w.out, 1, 'rec except uc1b').stdout).toMatch(/^APPLY:/m);
+    expect(answer(w.out, 1, recCarrying(w.out, 'rec', ['uc1b'])).stdout).toMatch(/^APPLY:/m);
     const withdraw = r.ap(['revise', '--out', w.out, '--withdraw', `${w.run}-uc1`]);
     expect(withdraw.status, withdraw.stderr).toBe(0);
     expect(withdraw.stdout).toMatch(/^APPROVED: gate_rev=1 \(every accepted change is in the plan or settled\)$/m);
@@ -179,7 +179,7 @@ describe('record-only changes settle without a plan edit', () => {
   test('a packaging change: APPLY says settle, revise --settle reopens nothing and completes, export and validate pass; a scope change never settles', () => {
     const g = gated(PACKAGING, {});
     expect(gate(g.out).items.find((i: any) => i.id === 'uc1')).toMatchObject({ recommended: 'b', changes: 'packaging' });
-    const res = answer(g.out, 1, 'rec');
+    const res = answer(g.out, 1, recCarrying(g.out, 'rec'));
     expect(res.status, res.stderr).toBe(0);
     expect(res.stdout).toMatch(new RegExp(`^APPLY: ${esc(g.run)}-uc1 .*; no plan edit needed: revise --settle uc1$`, 'm'));
     expect(readRunJson(g.out).status).toBe('amend_pending');
@@ -196,7 +196,7 @@ describe('record-only changes settle without a plan edit', () => {
     expect(r.ap(['export', '--out', g.out]).status).toBe(0);
     expect(r.validate(g.out).status).toBe(0);
     const scope = gated(AGREE, {});
-    expect(answer(scope.out, 1, 'rec except uc1b').stdout).toMatch(/; edit the plan, then revise$/m);
+    expect(answer(scope.out, 1, recCarrying(scope.out, 'rec', ['uc1b'])).stdout).toMatch(/; edit the plan, then revise$/m);
     const refused = r.ap(['revise', '--out', scope.out, '--settle', 'uc1']);
     expect(refused.status).toBe(3);
     expect(refused.stderr).toMatch(/changes scope; only packaging or ordering changes settle.*\(AMENDMENT_NOT_SETTLEABLE\)$/m);
@@ -229,7 +229,7 @@ describe('mixed amendments: a scope change and a packaging change', () => {
     expect(readRunJson(g.out).status).toBe('running');
     phaseVia(r, g.out, 'ceo');
     expect(phaseVia(r, g.out, 'eng').outside.stdout).toContain('status=gate_pending');
-    expect(answer(g.out, 2, 'rec').stdout).toContain('status=complete');
+    expect(answer(g.out, 2, recCarrying(g.out, 'rec')).stdout).toContain('status=complete');
     expect(r.validate(g.out).status).toBe(0);
   });
   test('settle and incorporate in one revise; settling alone on an unreviewed plan edit is refused', () => {
@@ -262,7 +262,7 @@ describe('partial replies resolve against persisted decisions', () => {
 
   test('validate rejects a complete run with a pending approval', () => {
     const g = gated({}, {});
-    expect(answer(g.out, 1, 'all').stdout).toContain('status=complete');
+    expect(answer(g.out, 1, recCarrying(g.out)).stdout).toContain('status=complete');
     const file = path.join(g.out, 'decisions.jsonl');
     fs.writeFileSync(file, decisions(g.out).map(d => JSON.stringify(d.label === 'UC2' ? { ...d, status: 'pending', chosen: null } : d)).join('\n') + '\n');
     r.ap(['export', '--out', g.out]);
@@ -298,3 +298,39 @@ describe('reruns', () => {
     expect(decisions(g.out).find(d => d.gate_rev === 2 && d.label === demoted.id.toUpperCase())).toMatchObject({ chosen: 'b', source: 'agent' });
   });
 });
+
+describe('open Critical and High findings reach the gate (GBRA-223 claim 3)', () => {
+  const NONE = { native: [], outside: [] };
+  const KILL = { native: [{ id: 'F1', severity: 'Critical', title: 'No kill switch for the resolver rollout', fix: 'add a config kill switch' }], outside: [] };
+  test('each open Critical/High finding is an apply item that rec takes; a twin pair is one item; Medium stays a task', () => {
+    const g = gated({
+      native: [{ id: 'F1', severity: 'Critical', title: 'No kill switch for the resolver rollout', fix: 'add a config kill switch' }, { id: 'F2', severity: 'Medium', title: 'Name the flag consistently' }],
+      outside: [{ id: 'O1', severity: 'High', title: 'No kill switch for the resolver rollout', fix: 'ship behind a kill switch' }],
+    }, { native: [{ id: 'F3', severity: 'High', title: 'No test covers a hung resolver', file: 'test/resolver.test.ts', fix: 'add a hung-resolver test' }], outside: [] });
+    const items = gate(g.out).items;
+    const apply = items.filter((i: any) => i.finding);
+    expect(apply.map((i: any) => i.title)).toEqual([
+      expect.stringMatching(/^Critical: No kill switch for the resolver rollout \[.*-ceo-native-1 \+ .*-ceo-outside-1\]$/),
+      expect.stringMatching(/^High: No test covers a hung resolver \[.*-eng-native-1\]$/),
+    ]);
+    for (const i of apply) expect(i).toMatchObject({ kind: 'auto', recommended: 'a', options: [{ key: 'a', changes_plan: true }, { key: 'b', changes_plan: false }] });
+    expect(apply[0].options[0].text).toContain('add a config kill switch');
+    expect(apply[0].options[0].text).toContain('ship behind a kill switch');
+    expect(items.find((i: any) => i.id === 'p1')).toMatchObject({ recommended: 'b' });
+    expect(items.some((i: any) => /Name the flag/.test(i.title))).toBe(false);
+    const rec = answer(g.out, 1, 'rec');
+    expect(rec.stdout.match(/^APPLY: /gm)?.length).toBe(2);
+    expect(readRunJson(g.out).status).toBe('amend_pending');
+  });
+
+  test('b, as-is, or a token reply that leaves the item unanswered carries the finding into the build with no plan change', () => {
+    for (const pick of [(id: string) => `${id}b p1a`, () => 'as-is', () => 'p1a']) {
+      const g = gated(KILL, NONE);
+      const item = gate(g.out).items.find((i: any) => i.finding);
+      const done = answer(g.out, 1, pick(item.id), true);
+      expect(JSON.parse(done.stdout)).toMatchObject({ outcome: 'complete', reopened: [], amendments: [] });
+      expect(decisions(g.out).find(d => d.label === item.id.toUpperCase())).toMatchObject({ chosen: 'b' });
+      expect(readJsonl(path.join(g.out, 'tasks.jsonl')).some(t => /No kill switch/.test(t.title))).toBe(true);
+    }
+  });
+});

test/autoplan-run-recovery.test.ts

@@ -106,7 +106,8 @@ describe('next/submit loop', () => {
     expect(v.status, v.stdout + v.stderr).toBe(0);
     expect(JSON.parse(v.stdout)).toMatchObject({ valid: true, status: 'gate_pending' });
     const gate = JSON.parse(fs.readFileSync(path.join(out, 'gate.json'), 'utf8'));
-    expect(gate.items.map((i: any) => i.id)).toEqual(['uc1', 'uc2', 'p1']);
+    expect(gate.items.map((i: any) => i.id)).toEqual(['uc1', 'd1', 'd2', 'uc2', 'd3', 'd4', 'p1']);
+    expect(gate.items.filter((i: any) => i.finding).map((i: any) => i.title.split(':')[0])).toEqual(['Critical', 'High', 'Critical', 'High']);
     const decisions = fs.readFileSync(path.join(out, 'decisions.jsonl'), 'utf8').trim().split('\n').map(l => JSON.parse(l));
     expect(decisions.every((d: any) => d.status === 'pending' && d.gate_rev === 1)).toBe(true);
     const tasks = fs.readFileSync(path.join(out, 'tasks.jsonl'), 'utf8').trim().split('\n').map(l => JSON.parse(l));

test/helpers/autoplan-run-fixture.ts

@@ -58,6 +58,16 @@ export function resultOf(out: string, phase: string, voice: 'native' | 'outside'
 }
 /** The sealed copy a completed voice is bound to. */
 export const boundOf = (out: string, phase: string, voice: 'native' | 'outside'): string => path.join(out, readRun(out).phases[phase][voice].output);
+/**
+ * A `rec`/`all` reply that carries every open Critical/High finding item into the build (`<id>b`, no plan change),
+ * so a test about another gate mechanism keeps its flow; `extra` tokens are appended to the except list.
+ */
+export function recCarrying(out: string, base = 'all', extra: string[] = []): string {
+  const carry = (JSON.parse(fs.readFileSync(path.join(out, 'gate.json'), 'utf8')).items as Array<{ id: string; finding?: string[] }>).filter(i => i.finding).map(i => `${i.id}b`);
+  const tokens = [...carry, ...extra];
+  return tokens.length ? `${base} except ${tokens.join(' ')}` : base;
+}
+
 export function attemptsOf(out: string, phase: string): { native: string; outside: string } {
   const run = readRun(out);
   return { native: run.phases[phase].native.attempt, outside: run.phases[phase].outside.attempt };