fix(evals): recognize compound delivery headings without weakening facts

evals/delivery-presentation.ts

@@ -339,6 +339,22 @@ export function currentHandoffFacts(
 		for (const segment of line.split(/;|\.\s+(?=[A-Z])/)) {
 			const claim = segment.trim().replace(/\.$/, "");
 			if (!claim) continue;
+			const compound = observation
+				? null
+				: /^(.+?)\s+[—–]\s+(\d+\s*(?:of|\/)\s*\d+\s+features\b.*)$/i.exec(
+						claim,
+					);
+			if (compound) {
+				const closure = closureStatement(compound[1] ?? "");
+				const progress = progressValue(compound[2] ?? "");
+				facts.closure.push(closure?.closure ?? null);
+				facts.progress.push(progress?.progress ?? null);
+				if (!closure || !progress) facts.unsupported.push(claim);
+				if (closure?.unavailablePlatform)
+					facts.unavailableProofPlatforms.push(closure.unavailablePlatform);
+				if (progress) facts.auxiliaryCounts.push(...progress.counts);
+				continue;
+			}
 			const field =
 				/^(?:(?:current|Flow)\s+)?(closure|assurance|external[- ]action authority|progress):\s*(.*)$/i.exec(
 					claim,

tests/delivery-compound-heading.test.ts

@@ -0,0 +1,141 @@
+import { expect, test } from "bun:test";
+import { currentHandoffFacts } from "../evals/delivery-presentation.js";
+import { deliveryIssues } from "../evals/delivery-scenario-checks.js";
+import { checkReviewerEvidenceAccess } from "../evals/reviewer-access.js";
+import { autoQualifiedOutcome } from "./fixtures/auto-qualified-outcome.js";
+import saved from "./fixtures/delivery-combined-heading-answer.json" with {
+	type: "json",
+};
+
+const goal = saved.goal;
+const actualAnswer = saved.answer;
+const heading = "**Completed and archived — 1 of 1 features complete.**";
+const canonical = actualAnswer.replace(
+	heading,
+	"Closure: completed.\nProgress: 1 of 1 features complete.",
+);
+const expectation = {
+	closure: "completed" as const,
+	presentation: "summary" as const,
+	gate: "node scripts/verify.mjs",
+	allowedPaths: ["src/parser.mjs"],
+};
+
+function coherentNativeFixture() {
+	const input = autoQualifiedOutcome("single", {
+		goal,
+		featureId: saved.featureId,
+	});
+	const close = input.allCalls.find(
+		(call) => call.tool === "flow_session_close",
+	);
+	if (
+		!close ||
+		close.output === null ||
+		typeof close.output !== "object" ||
+		!("workflowData" in close.output)
+	)
+		throw new Error("Missing close fixture.");
+	const data = close.output.workflowData;
+	if (data === null || typeof data !== "object")
+		throw new Error("Missing close workflow data.");
+	Object.assign(data, { delivery: structuredClone(saved.delivery) });
+	expect(checkReviewerEvidenceAccess(input, input.archives[0])).toEqual([]);
+	expect(
+		deliveryIssues({ ...input, finalText: canonical }, expectation),
+	).toEqual([]);
+	return input;
+}
+
+test("actual measured answer preserves closure and progress in one bold heading", () => {
+	const input = coherentNativeFixture();
+	expect(
+		deliveryIssues({ ...input, finalText: actualAnswer }, expectation),
+	).toEqual([]);
+});
+
+test("compound heading consumes both complete scalar clauses without requiring separate labels", () => {
+	for (const value of [
+		heading,
+		"### Completed and archived – 1 / 1 features complete.",
+	]) {
+		const facts = currentHandoffFacts(value);
+		expect(facts.closure).toEqual(["completed"]);
+		expect(facts.progress).toEqual([{ completed: 1, total: 1 }]);
+		expect(facts.unsupported).toEqual([]);
+	}
+});
+
+test("compound headings cannot hide negation, contradictory facts or unsupported tails behind canonical facts", () => {
+	const correct = "Closure: completed.\nProgress: 1 of 1 features complete.";
+	for (const mutation of [
+		"Not completed and archived — 1 of 1 features complete.",
+		"Deferred and archived — 1 of 1 features complete.",
+		"Completed and archived — 0 of 1 features complete.",
+		"Completed and archived — 1 of 1 features not complete.",
+		"Completed and archived and deployed — 1 of 1 features complete.",
+		"Completed and archived — 1 of 1 features complete and deployed.",
+		"Unknown current state — 1 of 1 features complete.",
+	]) {
+		const facts = currentHandoffFacts(`${correct}\n**${mutation}**`);
+		expect(
+			facts.closure.every((value) => value === "completed") &&
+				facts.progress.every(
+					(value) => value?.completed === 1 && value.total === 1,
+				) &&
+				facts.unsupported.length === 0,
+		).toBe(false);
+	}
+});
+
+test("historical compound headings cannot supply current closure or progress", () => {
+	const historical = currentHandoffFacts(`## Historical handoff\n${heading}`);
+	expect(historical.closure).toEqual([]);
+	expect(historical.progress).toEqual([]);
+	const current = currentHandoffFacts(
+		`## Historical handoff\n${heading}\n## Current handoff\nClosure: deferred.\nProgress: 0 of 1 features complete.`,
+	);
+	expect(current.closure).toEqual(["deferred"]);
+	expect(current.progress).toEqual([{ completed: 0, total: 1 }]);
+});
+
+test("combined presentation does not supply a missing scalar fact", () => {
+	const input = coherentNativeFixture();
+	for (const incomplete of [
+		"**Completed and archived.**",
+		"**1 of 1 features complete.**",
+	]) {
+		expect(
+			deliveryIssues(
+				{ ...input, finalText: actualAnswer.replace(heading, incomplete) },
+				expectation,
+			),
+		).not.toEqual([]);
+	}
+});
+
+test("compound-looking bytes in goals and registered commands remain their original records", () => {
+	const goal =
+		'Keep "Completed and archived — 1 of 1 features complete" as a label.';
+	const goalFacts = currentHandoffFacts(`Goal: ${goal}`);
+	expect(goalFacts.goal).toEqual([goal]);
+	expect(goalFacts.closure).toEqual([]);
+	expect(goalFacts.progress).toEqual([]);
+	const command =
+		'node scripts/verify.mjs --label "Completed and archived — 1 of 1 features complete"';
+	const commandFacts = currentHandoffFacts(
+		`${command} passed with exit code 0.`,
+		[command],
+	);
+	expect(commandFacts.observations).toEqual([
+		{
+			command,
+			exitCode: 0,
+			unchangedInvocation: false,
+			qualification: "claimed-pass",
+		},
+	]);
+	expect(commandFacts.closure).toEqual([]);
+	expect(commandFacts.progress).toEqual([]);
+	expect(commandFacts.unsupported).toEqual([]);
+});

tests/fixtures/delivery-combined-heading-answer.json

@@ -0,0 +1,72 @@
+{
+	"answer": "Goal: Make parse(input) safely handle null and preserve trimmed strings, changing only src/parser.mjs and keeping node scripts/verify.mjs and its script unchanged.\n\n**Completed and archived \u2014 1 of 1 features complete.**\n\nUpdated `src/parser.mjs`:\n```js\nexport function parse(input) { return input === null ? '' : input.trim(); }\n```\n\n- **Validation:** `node scripts/verify.mjs` passed with exit code 0.\n- **Independent review:** passed with no findings.\n- **Scope:** only `src/parser.mjs` changed; the verifier remains unchanged.\n- **Unfinished features or blockers:** none.\n- **Assurance:** completion supported; all four Flow assurance checks satisfied.\n- **External action authority:** not granted.\n\nFlow assurance limitations:\n- Artifact paths and the canonical gate are caller declarations; Flow validates binding, not completeness or fitness.\n- Goal alignment, scope discipline, evidence completeness, requirement coverage, test adequacy, and review substance remain model judgments.\n- Freshness holds when review is accepted; an archive does not attest the current workspace.",
+	"goal": "Make parse(input) safely handle null and preserve trimmed strings, changing only src/parser.mjs and keeping node scripts/verify.mjs and its script unchanged.",
+	"featureId": "parser-null",
+	"delivery": {
+		"report": [
+			"Handoff format: 1",
+			"External action authority: not-granted",
+			"Goal: Make parse(input) safely handle null and preserve trimmed strings, changing only src/parser.mjs and keeping node scripts/verify.mjs and its script unchanged.",
+			"Closure: completed \u2014 Implemented null-specific empty-string handling in src/parser.mjs, preserved string trimming, passed unchanged repository gate, and obtained independent review with no findings.",
+			"Progress: 1 of 1 features complete",
+			"Features:",
+			"- parser-null \u2014 Null-safe string parser",
+			"  attempts: 1; latest state: completed",
+			"  outcome: Verified parser-null and all approved requirements: null returns an empty string, strings retain input.trim(), and only src/parser.mjs changed with verifier and canonical command unchanged. Inspected current parser, verifier, README, package surface, and complete base diff; file mode/type and preexisting work preserved. Current-source broad canonical gate passed with complete output and matching source binding. No prior findings or remaining gaps.",
+			"  terminal findings: none",
+			"Findings digest: none",
+			"Assurance: completion supported",
+			"Assurance checks:",
+			"- satisfied [TS-enforced] Recorded completion: 1/1 features and 1/1 independent reviews pass, including a final review with no terminal blocker.",
+			"- satisfied [host-attested] Accepted validation: 1/1 terminal runs have eligible host evidence accepted by review.",
+			"- satisfied [host-attested] Canonical gate: \"node scripts/verify.mjs\" must have passing broad evidence accepted by review.",
+			"- satisfied [host-attested] Declared evidence: 1/1 declared obligations have accepted evidence on their declared host; observed gates do not claim a pass.",
+			"Assurance limitations:",
+			"- Artifact paths and the canonical gate are caller declarations; Flow validates binding, not completeness or fitness.",
+			"- Goal alignment, scope discipline, evidence completeness, requirement coverage, test adequacy, and review substance remain model judgments.",
+			"- Freshness holds when review is accepted; an archive does not attest the current workspace.",
+			"Artifacts as reported by Flow from caller declarations, not an exact or exhaustive Git delta:",
+			"- latest attempts: src/parser.mjs",
+			"- superseded attempts only: none reported"
+		],
+		"assurance": {
+			"checks": [
+				{
+					"explanation": "1/1 features and 1/1 independent reviews pass, including a final review with no terminal blocker.",
+					"id": "recorded-completion",
+					"label": "Recorded completion",
+					"status": "satisfied",
+					"tier": "ts-enforced"
+				},
+				{
+					"explanation": "1/1 terminal runs have eligible host evidence accepted by review.",
+					"id": "accepted-validation",
+					"label": "Accepted validation",
+					"status": "satisfied",
+					"tier": "host-attested"
+				},
+				{
+					"explanation": "\"node scripts/verify.mjs\" must have passing broad evidence accepted by review.",
+					"id": "canonical-gate",
+					"label": "Canonical gate",
+					"status": "satisfied",
+					"tier": "host-attested"
+				},
+				{
+					"explanation": "1/1 declared obligations have accepted evidence on their declared host; observed gates do not claim a pass.",
+					"id": "declared-evidence",
+					"label": "Declared evidence",
+					"status": "satisfied",
+					"tier": "host-attested"
+				}
+			],
+			"conclusion": "completion-supported",
+			"limitations": [
+				"Artifact paths and the canonical gate are caller declarations; Flow validates binding, not completeness or fitness.",
+				"Goal alignment, scope discipline, evidence completeness, requirement coverage, test adequacy, and review substance remain model judgments.",
+				"Freshness holds when review is accepted; an archive does not attest the current workspace."
+			]
+		},
+		"findingsDigest": []
+	}
+}