omegon-pi 0.12.0

package/extensions/cleave/openspec.ts

@@ -0,0 +1,811 @@
+/**
+ * cleave/openspec — OpenSpec tasks.md parser.
+ *
+ * Parses OpenSpec's tasks.md format into ChildPlan[] for cleave execution.
+ * OpenSpec tasks.md uses numbered, grouped tasks with checkboxes:
+ *
+ *   ## 1. Theme Infrastructure
+ *   - [ ] 1.1 Create ThemeContext with light/dark state
+ *   - [ ] 1.2 Add CSS custom properties for colors
+ *
+ *   ## 2. UI Components
+ *   - [ ] 2.1 Create ThemeToggle component
+ *   - [ ] 2.2 Add toggle to settings page
+ *
+ * Each top-level group (## N. Title) becomes a ChildPlan.
+ * Subtasks within a group become the scope/description.
+ * Group ordering defines dependencies (later groups may depend on earlier).
+ */
+
+import { existsSync, readFileSync, readdirSync, writeFileSync, statSync } from "node:fs";
+import { join, basename } from "node:path";
+import type { ChildPlan, SplitPlan } from "./types.ts";
+
+// ─── Types ──────────────────────────────────────────────────────────────────
+
+export interface OpenSpecChange {
+	/** Change directory name (e.g., "add-dark-mode") */
+	name: string;
+	/** Full path to the change directory */
+	path: string;
+	/** Whether tasks.md exists */
+	hasTasks: boolean;
+	/** Whether proposal.md exists */
+	hasProposal: boolean;
+	/** Whether design.md exists */
+	hasDesign: boolean;
+}
+
+export interface TaskGroup {
+	/** Group number (1-based) */
+	number: number;
+	/** Group title (e.g., "Theme Infrastructure") */
+	title: string;
+	/** Individual tasks within the group */
+	tasks: Array<{
+		id: string;      // e.g., "1.1"
+		text: string;    // e.g., "Create ThemeContext with light/dark state"
+		done: boolean;   // checkbox state
+	}>;
+	/** Spec domains declared via <!-- specs: domain/name, ... --> annotation */
+	specDomains: string[];
+	/** Skill names declared via <!-- skills: skill1, skill2 --> annotation */
+	skills: string[];
+}
+
+/**
+ * Rich context extracted from an OpenSpec change, beyond just tasks.
+ * Carries design decisions, file scope, and spec scenarios for
+ * child enrichment and post-merge verification.
+ */
+export interface OpenSpecContext {
+	/** The change directory path */
+	changePath: string;
+	/** Full design.md content (null if absent) */
+	designContent: string | null;
+	/** Architecture decisions extracted from design.md */
+	decisions: string[];
+	/** Explicit file changes from design.md "File Changes" section */
+	fileChanges: Array<{ path: string; action: "new" | "modified" | "deleted" | "unknown" }>;
+	/** Delta spec scenarios for post-merge verification */
+	specScenarios: Array<{ domain: string; requirement: string; scenarios: string[] }>;
+	/** OpenAPI/AsyncAPI contract content (null if absent) */
+	apiContract: string | null;
+}
+
+// ─── Detection ──────────────────────────────────────────────────────────────
+
+/**
+ * Detect whether an OpenSpec workspace exists in the given repo.
+ * Returns the path to openspec/ if found, null otherwise.
+ */
+export function detectOpenSpec(repoPath: string): string | null {
+	const openspecDir = join(repoPath, "openspec");
+	return existsSync(openspecDir) ? openspecDir : null;
+}
+
+/**
+ * List active (non-archived) OpenSpec changes.
+ */
+export function listChanges(openspecDir: string): OpenSpecChange[] {
+	const changesDir = join(openspecDir, "changes");
+	if (!existsSync(changesDir)) return [];
+
+	const entries = readdirSync(changesDir, { withFileTypes: true });
+	const changes: OpenSpecChange[] = [];
+
+	for (const entry of entries) {
+		if (!entry.isDirectory() || entry.name === "archive") continue;
+
+		const changePath = join(changesDir, entry.name);
+		changes.push({
+			name: entry.name,
+			path: changePath,
+			hasTasks: existsSync(join(changePath, "tasks.md")),
+			hasProposal: existsSync(join(changePath, "proposal.md")),
+			hasDesign: existsSync(join(changePath, "design.md")),
+		});
+	}
+
+	return changes;
+}
+
+/**
+ * Find changes that have tasks.md ready for execution.
+ */
+export function findExecutableChanges(openspecDir: string): OpenSpecChange[] {
+	return listChanges(openspecDir).filter((c) => c.hasTasks);
+}
+
+// ─── Parsing ────────────────────────────────────────────────────────────────
+
+/**
+ * Parse an OpenSpec tasks.md into task groups.
+ *
+ * Supports formats:
+ *   ## 1. Group Title
+ *   - [ ] 1.1 Task description
+ *   - [x] 1.2 Completed task
+ *
+ * Also handles unnumbered groups:
+ *   ## Group Title
+ *   - [ ] Task description
+ */
+export function parseTasksFile(content: string): TaskGroup[] {
+	const groups: TaskGroup[] = [];
+	let currentGroup: TaskGroup | null = null;
+
+	const lines = content.split("\n");
+
+	for (const line of lines) {
+		// Match group header: ## 1. Title or ## Title
+		const groupMatch = line.match(/^##\s+(?:(\d+)\.\s+)?(.+)$/);
+		if (groupMatch) {
+			if (currentGroup) groups.push(currentGroup);
+			currentGroup = {
+				number: groupMatch[1] ? parseInt(groupMatch[1], 10) : groups.length + 1,
+				title: groupMatch[2].trim(),
+				tasks: [],
+				specDomains: [],
+				skills: [],
+			};
+			continue;
+		}
+
+		// Match spec-domain annotation: <!-- specs: domain/name, domain2/name2 -->
+		const specMatch = line.match(/^\s*<!--\s*specs:\s*(.+?)\s*-->\s*$/);
+		if (specMatch && currentGroup && currentGroup.tasks.length === 0) {
+			currentGroup.specDomains = specMatch[1]
+				.split(",")
+				.map((s) => s.trim())
+				.filter((s) => s.length > 0);
+			continue;
+		}
+
+		// Match skills annotation: <!-- skills: python, k8s-operations -->
+		const skillsMatch = line.match(/^\s*<!--\s*skills:\s*(.+?)\s*-->\s*$/);
+		if (skillsMatch && currentGroup && currentGroup.tasks.length === 0) {
+			currentGroup.skills = skillsMatch[1]
+				.split(",")
+				.map((s) => s.trim())
+				.filter((s) => s.length > 0);
+			continue;
+		}
+
+		// Match task item: - [ ] 1.1 Description or - [x] 1.2 Description
+		const taskMatch = line.match(/^\s*-\s+\[([ xX])\]\s+(?:(\d+(?:\.\d+)?)\s+)?(.+)$/);
+		if (taskMatch && currentGroup) {
+			currentGroup.tasks.push({
+				id: taskMatch[2] || `${currentGroup.number}.${currentGroup.tasks.length + 1}`,
+				text: taskMatch[3].trim(),
+				done: taskMatch[1] !== " ",
+			});
+			continue;
+		}
+
+		// Match unnumbered bullet task under a group: - Task text (no checkbox)
+		const bulletMatch = line.match(/^\s*-\s+(?!\[)(.+)$/);
+		if (bulletMatch && currentGroup) {
+			currentGroup.tasks.push({
+				id: `${currentGroup.number}.${currentGroup.tasks.length + 1}`,
+				text: bulletMatch[1].trim(),
+				done: false,
+			});
+		}
+	}
+
+	if (currentGroup) groups.push(currentGroup);
+	return groups;
+}
+
+// ─── Conversion ─────────────────────────────────────────────────────────────
+
+/**
+ * Convert OpenSpec task groups to cleave ChildPlan[].
+ *
+ * Each group becomes a child. Dependencies are inferred from:
+ * - Explicit markers in title: "after X", "requires X", "depends on X"
+ * - Task text references to earlier group titles
+ *
+ * Groups where ALL tasks are already done are filtered out.
+ *
+ * Returns null if fewer than 2 executable groups (not worth cleaving).
+ */
+export function taskGroupsToChildPlans(groups: TaskGroup[]): ChildPlan[] | null {
+	// Filter out groups where all tasks are done
+	const activeGroups = groups.filter((g) =>
+		g.tasks.length === 0 || g.tasks.some((t) => !t.done),
+	);
+
+	if (activeGroups.length < 2) return null;
+
+	// Cap at 4 children (cleave limit)
+	const effectiveGroups = activeGroups.length > 4 ? mergeSmallGroups(activeGroups, 4) : activeGroups;
+
+	const plans: ChildPlan[] = effectiveGroups.map((group) => {
+		const label = group.title
+			.toLowerCase()
+			.replace(/[^\w\s-]/g, "")
+			.replace(/[\s_]+/g, "-")
+			.replace(/-+/g, "-")
+			.replace(/^-|-$/g, "")
+			.slice(0, 40);
+
+		const taskDescriptions = group.tasks
+			.filter((t) => !t.done) // Skip already-completed tasks
+			.map((t) => `- ${t.text}`);
+
+		const description = taskDescriptions.length > 0
+			? `${group.title}:\n${taskDescriptions.join("\n")}`
+			: group.title;
+
+		// Infer scope from task text: look for file paths and patterns
+		const scope = inferScope(group.tasks.map((t) => t.text));
+
+		return {
+			label,
+			description,
+			scope,
+			dependsOn: [] as string[],
+			specDomains: [...(group.specDomains ?? [])],
+			skills: [...(group.skills ?? [])],
+		};
+	});
+
+	// Infer dependencies from explicit markers and title references
+	inferDependencies(plans);
+
+	return plans;
+}
+
+/**
+ * Infer inter-group dependencies from explicit markers in descriptions.
+ *
+ * Looks for patterns like:
+ * - "after <label>" or "after <title words>"
+ * - "requires <label>"
+ * - "depends on <label>"
+ * - Task text referencing an earlier group's title
+ */
+function inferDependencies(plans: ChildPlan[]): void {
+	const labelSet = new Set(plans.map((p) => p.label));
+
+	for (let i = 0; i < plans.length; i++) {
+		const text = plans[i].description.toLowerCase();
+
+		for (let j = 0; j < plans.length; j++) {
+			if (i === j) continue;
+			const otherLabel = plans[j].label;
+			// Convert label back to words for fuzzy matching: "database-layer" → "database layer"
+			const otherWords = otherLabel.replace(/-/g, " ");
+
+			// Explicit markers: "after X", "requires X", "depends on X"
+			const markers = [
+				`after ${otherLabel}`, `after ${otherWords}`,
+				`requires ${otherLabel}`, `requires ${otherWords}`,
+				`depends on ${otherLabel}`, `depends on ${otherWords}`,
+			];
+
+			if (markers.some((m) => text.includes(m))) {
+				if (labelSet.has(otherLabel) && !plans[i].dependsOn.includes(otherLabel)) {
+					plans[i].dependsOn.push(otherLabel);
+				}
+			}
+		}
+	}
+}
+
+/**
+ * Full pipeline: read an OpenSpec change and convert to SplitPlan.
+ *
+ * Returns null if the change doesn't have tasks or has fewer than 2 groups.
+ */
+export function openspecChangeToSplitPlan(changePath: string): SplitPlan | null {
+	const tasksPath = join(changePath, "tasks.md");
+	if (!existsSync(tasksPath)) return null;
+
+	const content = readFileSync(tasksPath, "utf-8");
+	const groups = parseTasksFile(content);
+	const children = taskGroupsToChildPlans(groups);
+	if (!children) return null;
+
+	// Read proposal for rationale if available
+	let rationale = `From OpenSpec change: ${basename(changePath)}`;
+	const proposalPath = join(changePath, "proposal.md");
+	if (existsSync(proposalPath)) {
+		const proposal = readFileSync(proposalPath, "utf-8");
+		// Extract intent section
+		const intentMatch = proposal.match(/##\s+Intent\s*\n([\s\S]*?)(?=\n##|$)/);
+		if (intentMatch) {
+			rationale = intentMatch[1].trim().slice(0, 200);
+		}
+	}
+
+	return { children, rationale };
+}
+
+// ─── Design Context ─────────────────────────────────────────────────────────
+
+/**
+ * Parse the "File Changes" section from design.md.
+ *
+ * Supports formats:
+ *   - `src/contexts/ThemeContext.tsx` (new)
+ *   - `src/styles/globals.css` (modified)
+ *   - src/old/file.ts (deleted)
+ *   - `path/to/file.ts`  (no action → unknown)
+ */
+export function parseDesignFileChanges(
+	designContent: string,
+): Array<{ path: string; action: "new" | "modified" | "deleted" | "unknown" }> {
+	const results: Array<{ path: string; action: "new" | "modified" | "deleted" | "unknown" }> = [];
+
+	// Find the File Changes section
+	const sectionMatch = designContent.match(
+		/##\s+File\s+Changes?\s*\n([\s\S]*?)(?=\n##\s|\n#\s|$)/i,
+	);
+	if (!sectionMatch) return results;
+
+	const section = sectionMatch[1];
+	// Match lines like: - `path/to/file` (action)  or  - path/to/file (action)
+	const lineRe = /^[\s-]*[`"']?([a-zA-Z0-9_./-]+\.[a-zA-Z0-9]+)[`"']?\s*(?:\((\w+)\))?/gm;
+	let m: RegExpExecArray | null;
+	while ((m = lineRe.exec(section)) !== null) {
+		const filePath = m[1];
+		const rawAction = (m[2] || "").toLowerCase();
+		let action: "new" | "modified" | "deleted" | "unknown" = "unknown";
+		if (rawAction === "new" || rawAction === "created" || rawAction === "create") action = "new";
+		else if (rawAction === "modified" || rawAction === "updated" || rawAction === "modify") action = "modified";
+		else if (rawAction === "deleted" || rawAction === "removed" || rawAction === "delete") action = "deleted";
+		results.push({ path: filePath, action });
+	}
+
+	return results;
+}
+
+/**
+ * Extract architecture decisions from design.md.
+ *
+ * Looks for "### Decision:" headers and captures the title + rationale.
+ */
+export function parseDesignDecisions(designContent: string): string[] {
+	const decisions: string[] = [];
+	const re = /###\s+Decision:\s*(.+?)(?:\n[\s\S]*?(?=\n###|\n##|$))/g;
+	let m: RegExpExecArray | null;
+	while ((m = re.exec(designContent)) !== null) {
+		// Capture the decision title and first line of rationale
+		const title = m[0];
+		const lines = title.split("\n").filter((l) => l.trim());
+		const summary = lines.length > 1
+			? `${lines[0].replace(/^###\s+Decision:\s*/, "").trim()}: ${lines.slice(1).find((l) => !l.startsWith("#"))?.trim() || ""}`
+			: lines[0].replace(/^###\s+Decision:\s*/, "").trim();
+		decisions.push(summary);
+	}
+	return decisions;
+}
+
+// ─── Spec Scenarios ─────────────────────────────────────────────────────────
+
+/**
+ * Read delta spec files from a change and extract scenarios for verification.
+ *
+ * Parses Given/When/Then scenarios from ADDED and MODIFIED requirements
+ * in the change's specs/ directory.
+ */
+export function readSpecScenarios(
+	changePath: string,
+): Array<{ domain: string; requirement: string; scenarios: string[] }> {
+	const specsDir = join(changePath, "specs");
+	if (!existsSync(specsDir)) return [];
+
+	const results: Array<{ domain: string; requirement: string; scenarios: string[] }> = [];
+
+	// Recursively find spec.md files
+	const specFiles = findSpecFiles(specsDir);
+
+	for (const specFile of specFiles) {
+		const content = readFileSync(specFile, "utf-8");
+		const domain = specFile
+			.replace(specsDir + "/", "")
+			.replace(/\/spec\.md$/, "")
+			.replace(/\.md$/, "");
+
+		// Only extract from ADDED and MODIFIED sections (these need verification)
+		const relevantSections = content.match(
+			/##\s+(?:ADDED|MODIFIED)\s+Requirements?\s*\n([\s\S]*?)(?=\n##\s+(?:ADDED|MODIFIED|REMOVED)|$)/gi,
+		);
+		if (!relevantSections) continue;
+
+		for (const section of relevantSections) {
+			// Find requirements with scenarios
+			const reqRe = /###\s+Requirement:\s*(.+)/g;
+			let reqMatch: RegExpExecArray | null;
+			while ((reqMatch = reqRe.exec(section)) !== null) {
+				const reqName = reqMatch[1].trim();
+				// Find scenarios after this requirement until next requirement or section end
+				const afterReq = section.slice(reqMatch.index + reqMatch[0].length);
+				const nextReq = afterReq.search(/\n###\s+Requirement:/);
+				const scenarioBlock = nextReq >= 0 ? afterReq.slice(0, nextReq) : afterReq;
+
+				const scenarios: string[] = [];
+				const scenarioRe = /####\s+Scenario:\s*(.+?)(?:\n[\s\S]*?)(?=\n####|\n###|$)/g;
+				let scenMatch: RegExpExecArray | null;
+				while ((scenMatch = scenarioRe.exec(scenarioBlock)) !== null) {
+					// Extract the full scenario including Given/When/Then
+					const scenarioText = scenMatch[0]
+						.replace(/^####\s+Scenario:\s*/, "")
+						.trim();
+					scenarios.push(scenarioText);
+				}
+
+				if (scenarios.length > 0) {
+					results.push({ domain, requirement: reqName, scenarios });
+				}
+			}
+		}
+	}
+
+	return results;
+}
+
+/** Recursively find spec.md files under a directory. */
+function findSpecFiles(dir: string): string[] {
+	const files: string[] = [];
+	if (!existsSync(dir)) return files;
+
+	const entries = readdirSync(dir, { withFileTypes: true });
+	for (const entry of entries) {
+		const fullPath = join(dir, entry.name);
+		if (entry.isDirectory()) {
+			files.push(...findSpecFiles(fullPath));
+		} else if (entry.name.endsWith(".md")) {
+			files.push(fullPath);
+		}
+	}
+	return files;
+}
+
+// ─── Full Context ───────────────────────────────────────────────────────────
+
+/**
+ * Build full OpenSpec context from a change directory.
+ *
+ * Reads design.md (decisions, file changes), delta specs (scenarios),
+ * and returns a structured context object that cleave uses to:
+ * - Enrich child task files with design context
+ * - Supply exact file scope from design file changes
+ * - Verify implementation against spec scenarios post-merge
+ */
+export function buildOpenSpecContext(changePath: string): OpenSpecContext {
+	const ctx: OpenSpecContext = {
+		changePath,
+		designContent: null,
+		decisions: [],
+		fileChanges: [],
+		specScenarios: [],
+		apiContract: null,
+	};
+
+	// Design
+	const designPath = join(changePath, "design.md");
+	if (existsSync(designPath)) {
+		ctx.designContent = readFileSync(designPath, "utf-8");
+		ctx.decisions = parseDesignDecisions(ctx.designContent);
+		ctx.fileChanges = parseDesignFileChanges(ctx.designContent);
+	}
+
+	// Specs
+	ctx.specScenarios = readSpecScenarios(changePath);
+
+	// API contract (OpenAPI / AsyncAPI) — check common extensions
+	for (const name of ["api.yaml", "api.yml", "api.json"]) {
+		const apiPath = join(changePath, name);
+		if (existsSync(apiPath)) {
+			ctx.apiContract = readFileSync(apiPath, "utf-8");
+			break;
+		}
+	}
+
+	return ctx;
+}
+
+/**
+ * Full pipeline: read an OpenSpec change and convert to SplitPlan + context.
+ *
+ * Returns null if the change doesn't have tasks or has fewer than 2 groups.
+ */
+export function openspecChangeToSplitPlanWithContext(
+	changePath: string,
+): { plan: SplitPlan; context: OpenSpecContext } | null {
+	const plan = openspecChangeToSplitPlan(changePath);
+	if (!plan) return null;
+
+	const context = buildOpenSpecContext(changePath);
+
+	// Supplement scope from design.md file changes when available
+	if (context.fileChanges.length > 0) {
+		supplementScopeFromDesign(plan.children, context.fileChanges);
+	}
+
+	return { plan, context };
+}
+
+/**
+ * Supplement child scope with explicit file changes from design.md.
+ *
+ * For each child, if design.md lists files that match the child's description
+ * or existing scope patterns, add them. This replaces heuristic guessing
+ * with author-declared intent.
+ */
+function supplementScopeFromDesign(
+	children: ChildPlan[],
+	fileChanges: Array<{ path: string; action: string }>,
+): void {
+	// If there's only one group of files, distribute to the closest-matching child
+	// If files are clearly separated by directory, match by path prefix
+
+	const filePaths = fileChanges
+		.filter((f) => f.action !== "deleted")
+		.map((f) => f.path);
+
+	if (filePaths.length === 0) return;
+
+	for (const child of children) {
+		const descLower = child.description.toLowerCase();
+		const labelWords = child.label.replace(/-/g, " ").split(" ");
+
+		const matched: string[] = [];
+		for (const fp of filePaths) {
+			const fpLower = fp.toLowerCase();
+			// Match if: file path contains a label word, or child description mentions the file
+			const pathParts = fpLower.split("/");
+			const isMatch =
+				labelWords.some((w) => w.length > 2 && pathParts.some((p) => p.includes(w))) ||
+				descLower.includes(fpLower) ||
+				child.scope.some((s) => {
+					const pattern = s.replace(/\*\*/g, "").replace(/\*/g, "");
+					return fpLower.startsWith(pattern) || pattern.startsWith(fpLower.split("/").slice(0, -1).join("/"));
+				});
+
+			if (isMatch) matched.push(fp);
+		}
+
+		// Add matched files to scope (deduplicated)
+		const existingScope = new Set(child.scope);
+		for (const fp of matched) {
+			if (!existingScope.has(fp)) {
+				child.scope.push(fp);
+			}
+		}
+	}
+}
+
+// ─── Helpers ────────────────────────────────────────────────────────────────
+
+/**
+ * Merge small groups to fit within maxGroups.
+ * Combines the smallest adjacent groups until we're at the limit.
+ */
+function mergeSmallGroups(groups: TaskGroup[], maxGroups: number): TaskGroup[] {
+	const result = [...groups];
+
+	while (result.length > maxGroups) {
+		// Find the smallest group by task count
+		let smallestIdx = 0;
+		let smallestSize = Infinity;
+		for (let i = 0; i < result.length - 1; i++) {
+			const combined = result[i].tasks.length + result[i + 1].tasks.length;
+			if (combined < smallestSize) {
+				smallestSize = combined;
+				smallestIdx = i;
+			}
+		}
+
+		// Merge with next group
+		const merged: TaskGroup = {
+			number: result[smallestIdx].number,
+			title: `${result[smallestIdx].title} + ${result[smallestIdx + 1].title}`,
+			tasks: [...result[smallestIdx].tasks, ...result[smallestIdx + 1].tasks],
+			specDomains: [
+				...(result[smallestIdx].specDomains ?? []),
+				...(result[smallestIdx + 1].specDomains ?? []),
+			],
+			skills: [
+				...new Set([
+					...(result[smallestIdx].skills ?? []),
+					...(result[smallestIdx + 1].skills ?? []),
+				]),
+			],
+		};
+		result.splice(smallestIdx, 2, merged);
+	}
+
+	return result;
+}
+
+// ─── Task Write-Back ────────────────────────────────────────────────────────
+
+/**
+ * After a successful cleave merge, mark completed child tasks as done
+ * in the original OpenSpec tasks.md.
+ *
+ * Maps completed child labels back to task groups and checks off their
+ * unchecked tasks. Returns the number of tasks marked done.
+ */
+export function writeBackTaskCompletion(
+	changePath: string,
+	completedLabels: string[],
+): { updated: number; totalTasks: number; allDone: boolean; unmatchedLabels: string[] } {
+	const tasksPath = join(changePath, "tasks.md");
+	if (!existsSync(tasksPath)) {
+		return { updated: 0, totalTasks: 0, allDone: false, unmatchedLabels: [...completedLabels] };
+	}
+
+	const content = readFileSync(tasksPath, "utf-8");
+	const groups = parseTasksFile(content);
+
+	// Build a set of completed label slugs for matching
+	const completedSet = new Set(completedLabels.map((l) => l.toLowerCase()));
+
+	// Track which group numbers are completed
+	const completedGroupNumbers = new Set<number>();
+	const matchedLabels = new Set<string>();
+	for (const group of groups) {
+		const groupSlug = group.title
+			.toLowerCase()
+			.replace(/[^\w\s-]/g, "")
+			.replace(/[\s_]+/g, "-")
+			.replace(/-+/g, "-")
+			.replace(/^-|-$/g, "")
+			.slice(0, 40);
+
+		if (completedSet.has(groupSlug)) {
+			completedGroupNumbers.add(group.number);
+			matchedLabels.add(groupSlug);
+		}
+	}
+
+	const unmatchedLabels = completedLabels
+		.map((label) => label.toLowerCase())
+		.filter((label) => !matchedLabels.has(label));
+
+	if (completedGroupNumbers.size === 0) {
+		const totalTasks = groups.reduce((sum, g) => sum + g.tasks.length, 0);
+		return { updated: 0, totalTasks, allDone: false, unmatchedLabels };
+	}
+
+	// Rewrite tasks.md line by line, checking off tasks in completed groups
+	const lines = content.split("\n");
+	let currentGroupNumber = -1;
+	let updated = 0;
+
+	for (let i = 0; i < lines.length; i++) {
+		// Detect group header
+		const groupMatch = lines[i].match(/^##\s+(?:(\d+)\.\s+)?(.+)$/);
+		if (groupMatch) {
+			currentGroupNumber = groupMatch[1] ? parseInt(groupMatch[1], 10) : -1;
+			// If unnumbered, find by title match
+			if (currentGroupNumber === -1) {
+				const title = groupMatch[2].trim();
+				const g = groups.find((g) => g.title === title);
+				if (g) currentGroupNumber = g.number;
+			}
+			continue;
+		}
+
+		// Check off unchecked tasks in completed groups
+		if (completedGroupNumbers.has(currentGroupNumber)) {
+			const taskMatch = lines[i].match(/^(\s*-\s+)\[ \](\s+.*)$/);
+			if (taskMatch) {
+				lines[i] = `${taskMatch[1]}[x]${taskMatch[2]}`;
+				updated++;
+			}
+		}
+	}
+
+	if (updated > 0) {
+		writeFileSync(tasksPath, lines.join("\n"), "utf-8");
+	}
+
+	// Check if all tasks are now done
+	const totalTasks = groups.reduce((sum, g) => sum + g.tasks.length, 0);
+	const wasDone = groups.reduce((sum, g) => sum + g.tasks.filter((t) => t.done).length, 0);
+	const allDone = wasDone + updated >= totalTasks;
+
+	return { updated, totalTasks, allDone, unmatchedLabels };
+}
+
+// ─── Active Changes Status ──────────────────────────────────────────────────
+
+export interface ChangeStatus {
+	name: string;
+	path: string;
+	totalTasks: number;
+	doneTasks: number;
+	hasProposal: boolean;
+	hasDesign: boolean;
+	hasSpecs: boolean;
+	/** Most recent mtime across change artifacts (ms since epoch) */
+	lastModifiedMs: number;
+}
+
+/**
+ * Summarize all active OpenSpec changes and their task completion status.
+ *
+ * Returns a list of changes with their task progress, suitable for
+ * session-start status display.
+ */
+export function getActiveChangesStatus(repoPath: string): ChangeStatus[] {
+	const openspecDir = detectOpenSpec(repoPath);
+	if (!openspecDir) return [];
+
+	const changes = listChanges(openspecDir);
+	const result: ChangeStatus[] = [];
+
+	for (const change of changes) {
+		let totalTasks = 0;
+		let doneTasks = 0;
+
+		if (change.hasTasks) {
+			const content = readFileSync(join(change.path, "tasks.md"), "utf-8");
+			const groups = parseTasksFile(content);
+			for (const group of groups) {
+				totalTasks += group.tasks.length;
+				doneTasks += group.tasks.filter((t) => t.done).length;
+			}
+		}
+
+		const specsDir = join(change.path, "specs");
+
+		// Find most recent modification time across artifacts
+		let lastModifiedMs = 0;
+		for (const file of ["tasks.md", "proposal.md", "design.md"]) {
+			const fp = join(change.path, file);
+			if (existsSync(fp)) {
+				try {
+					const mtime = statSync(fp).mtimeMs;
+					if (mtime > lastModifiedMs) lastModifiedMs = mtime;
+				} catch { /* skip */ }
+			}
+		}
+
+		result.push({
+			name: change.name,
+			path: change.path,
+			totalTasks,
+			doneTasks,
+			hasProposal: change.hasProposal,
+			hasDesign: change.hasDesign,
+			hasSpecs: existsSync(specsDir),
+			lastModifiedMs,
+		});
+	}
+
+	return result;
+}
+
+// ─── Helpers ────────────────────────────────────────────────────────────────
+
+/**
+ * Infer file scope patterns from task descriptions.
+ * Looks for quoted paths, file extensions, and common patterns.
+ */
+function inferScope(taskTexts: string[]): string[] {
+	const scope = new Set<string>();
+	const combined = taskTexts.join("\n");
+
+	// Backtick-quoted paths: `src/auth/login.ts`
+	for (const m of combined.matchAll(/`([a-zA-Z0-9_./-]+\.[a-zA-Z0-9]+)`/g)) {
+		scope.add(m[1]);
+	}
+
+	// Directory references: src/auth/, components/
+	for (const m of combined.matchAll(/\b((?:src|lib|app|components|pages|api|tests?|spec)\/?[a-zA-Z0-9_/-]*)\b/g)) {
+		const dir = m[1].replace(/\/$/, "");
+		if (dir.includes("/")) {
+			scope.add(dir + "/**");
+		}
+	}
+
+	return [...scope].slice(0, 10); // Cap at 10 patterns
+}