fix: add missing items for MCP array schemas

[that-lucas]

Summary

• Fix MCP tool schemas that defined type: 'array' without items, which breaks strict JSON Schema validation.
• This was reproduced in OpenCode: start a session with opus, then switch to codex; tool schema validation fails with array schema missing items.
• Add a regression test to scan src/mcp-tools/*.ts and fail if any array schema omits items.

Validation

• npx vitest run __tests__/mcp-schema-array-items.test.ts

Context

• Reported in Bug: MCP schema invalid for strict clients (array missing items) #1294 (OpenCode -> Codex session switch path).
• Closes Bug: MCP schema invalid for strict clients (array missing items) #1294

[Copilot]

Pull request overview

Fixes MCP JSON Schemas that define type: 'array' without items (which fails strict validation), and adds a regression test to prevent reintroducing the issue.

Changes:

• Added missing items definitions to several MCP tool input schemas that use type: 'array'.
• Added a Vitest regression test that scans MCP tool source files for array schemas missing items.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
v3/@claude-flow/cli/src/mcp-tools/system-tools.ts	Adds items to array-typed schema properties in system tools.
v3/@claude-flow/cli/src/mcp-tools/performance-tools.ts	Adds items to components array schema.
v3/@claude-flow/cli/src/mcp-tools/hooks-tools.ts	Adds items to trajectoryIds array schema.
v3/@claude-flow/cli/src/mcp-tools/github-tools.ts	Adds items to labels / assignees array schemas.
v3/@claude-flow/cli/src/mcp-tools/daa-tools.ts	Adds items to multiple array schemas and defines steps.items with anyOf.
v3/@claude-flow/cli/src/mcp-tools/coordination-tools.ts	Adds items to agents array schema.
v3/@claude-flow/cli/tests/mcp-schema-array-items.test.ts	Adds regression test to detect array schemas missing items.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

process.cwd() is not guaranteed to be v3/@claude-flow/cli when running Vitest in a monorepo/workspace, so this can fail with ENOENT depending on how tests are invoked. Prefer resolving the path relative to the test file (e.g., via import.meta.url + fileURLToPath and dirname) or via a known package root to make the test location-independent.

[Copilot]

The brace-matching approach can be thrown off by { / } that appear inside strings, template literals, or comments in the source (likely in large files like hooks-tools.ts), leading to false positives/negatives. Consider switching this check to a TypeScript AST-based approach (e.g., typescript parser) that specifically finds object literals where type is 'array' and verifies a sibling items property; this will be more robust and easier to evolve.

Suggested change

	function findArraySchemasMissingItems(source: string): number[] {
	const matches: number[] = [];
	const pattern = /type:\s*['\"]array['\"]/g;
	let match: RegExpExecArray | null;
	while ((match = pattern.exec(source)) !== null) {
	const typeIndex = match.index;
	const objectStart = source.lastIndexOf('{', typeIndex);
	if (objectStart === -1) {
	continue;
	}
	let depth = 0;
	let objectEnd = -1;
	for (let i = objectStart; i < source.length; i += 1) {
	const char = source[i];
	if (char === '{') {
	depth += 1;
	} else if (char === '}') {
	depth -= 1;
	if (depth === 0) {
	objectEnd = i;
	break;
	}
	}
	}
	if (objectEnd !== -1) {
	const objectText = source.slice(objectStart, objectEnd + 1);
	if (!/\bitems\s*:/.test(objectText)) {
	const line = source.slice(0, typeIndex).split('\n').length;
	matches.push(line);
	}
	}
	}
	return matches;
	import * as ts from 'typescript';
	function findArraySchemasMissingItems(source: string): number[] {
	const sourceFile = ts.createSourceFile(
	'file.ts',
	source,
	ts.ScriptTarget.Latest,
	true,
	ts.ScriptKind.TS
	);
	const lines: number[] = [];
	function checkObjectLiteral(node: ts.ObjectLiteralExpression) {
	let hasArrayType = false;
	let hasItems = false;
	let typePropNode: ts.PropertyAssignment | undefined;
	for (const prop of node.properties) {
	if (!ts.isPropertyAssignment(prop)) continue;
	const name = prop.name;
	const propName =
	ts.isIdentifier(name) || ts.isStringLiteralLike(name)
	? name.text
	: undefined;
	if (!propName) continue;
	if (propName === 'type') {
	if (
	ts.isStringLiteralLike(prop.initializer) &&
	prop.initializer.text === 'array'
	) {
	hasArrayType = true;
	typePropNode = prop;
	}
	} else if (propName === 'items') {
	hasItems = true;
	}
	}
	if (hasArrayType && !hasItems && typePropNode) {
	const { line } = sourceFile.getLineAndCharacterOfPosition(
	typePropNode.getStart(sourceFile)
	);
	lines.push(line + 1);
	}
	}
	function visit(node: ts.Node) {
	if (ts.isObjectLiteralExpression(node)) {
	checkObjectLiteral(node as ts.ObjectLiteralExpression);
	}
	ts.forEachChild(node, visit);
	}
	visit(sourceFile);
	return lines;

[Copilot]

Computing the line number by re-splitting the entire prefix on every match is O(n²) in the number of matches and file size (notable for large tool files). If this starts to slow the test suite, consider precomputing newline offsets once per file or counting newlines incrementally as the regex advances.