fix(tools): accept zero-argument OpenAI tool calls and reject non-object args

`getProfile`, `documentList` and `memoryForget` all declare `required: []`,
so a model may legitimately call them with no arguments at all. The OpenAI
API serialises that as `arguments: ""`, and `parseToolArguments` handed the
empty string straight to `JSON.parse`, so every no-argument call came back as

    {"success":false,"error":"Invalid JSON arguments for getProfile"}

Those three tools were unreachable in their documented no-argument form.

The same gate also lets non-object JSON through. `"null"` parses cleanly and
then rejects in the destructuring parameter of every tool function --
`TypeError: Cannot destructure property 'containerTag' of 'object null'` --
which escapes `executeToolCall`, since it has no catch, and fails the whole
request. That is precisely the throw #1488 added this gate to contain. `"5"`
and `"\"text\""` are quieter but worse: they destructure to `undefined` and
call the API with no container tag at all.

Treat blank arguments as `{}`, and require the parsed value to be a non-null,
non-array object. Malformed JSON still returns the tool error #1488 added.

Adds eight regression tests. Six of them fail against the current
implementation -- two on the blank-argument path and four on the non-object
path, one carrying the raw TypeError. The two guard tests, malformed JSON and
an ordinary well-formed call, pass both before and after, so the behaviour

packages/tools/src/openai/tools.ts

@@ -571,8 +571,21 @@ export function getToolDefinitions(): OpenAI.Chat.Completions.ChatCompletionTool
 }
 
 function parseToolArguments(argumentsJson: string) {
+	// getProfile, documentList and memoryForget all declare `required: []`, so a model
+	// may legitimately call them with no arguments. OpenAI serialises that as `""`,
+	// which is "no arguments" rather than malformed JSON — parse it as `{}`.
+	const source = argumentsJson?.trim() || "{}"
+
 	try {
-		return { success: true as const, value: JSON.parse(argumentsJson) }
+		const value = JSON.parse(source)
+
+		// `"null"`, `"5"` and `"[]"` parse cleanly, then throw in the destructuring
+		// parameter of every tool function — the throw this gate exists to contain.
+		if (typeof value !== "object" || value === null || Array.isArray(value)) {
+			return { success: false as const }
+		}
+
+		return { success: true as const, value }
 	} catch {
 		return { success: false as const }
 	}

packages/tools/src/tool-operations.test.ts

@@ -370,6 +370,87 @@ describe("memoryForget", () => {
 	})
 })
 
+describe("openai executeToolCall argument parsing", () => {
+	type ExecutorToolCall = Parameters<
+		ReturnType<typeof openAi.createToolCallExecutor>
+	>[0]
+
+	function toolCall(name: string, args: string) {
+		return {
+			id: "call_1",
+			type: "function",
+			function: { name, arguments: args },
+		} as ExecutorToolCall
+	}
+
+	// getProfile, documentList and memoryForget declare `required: []`, so the model
+	// is allowed to call them with no arguments. OpenAI serialises that as "".
+	it.each([
+		"",
+		"   ",
+	])("runs a zero-argument tool when arguments are %p", async (args) => {
+		const execute = openAi.createToolCallExecutor(API_KEY, {
+			containerTags: ["user_1"],
+		})
+
+		const result = JSON.parse(await execute(toolCall("getProfile", args)))
+
+		expect(result.success).toBe(true)
+		expect(profileRequest).toHaveBeenCalledTimes(1)
+		expect(profileRequest).toHaveBeenCalledWith({ containerTag: "user_1" })
+	})
+
+	// These parse cleanly, so the JSON guard lets them through to a destructuring
+	// parameter that rejects — the throw the guard exists to contain.
+	it.each([
+		"null",
+		"5",
+		"[]",
+		'"text"',
+	])("rejects non-object arguments %p as a tool result rather than throwing", async (args) => {
+		const execute = openAi.createToolCallExecutor(API_KEY)
+
+		const result = JSON.parse(await execute(toolCall("getProfile", args)))
+
+		expect(result.success).toBe(false)
+		expect(result.error).toMatch(/Invalid JSON arguments for getProfile/)
+		expect(profileRequest).not.toHaveBeenCalled()
+	})
+
+	it("still reports malformed JSON as a tool error", async () => {
+		const execute = openAi.createToolCallExecutor(API_KEY)
+
+		const result = JSON.parse(
+			await execute(toolCall("searchMemories", "{not json")),
+		)
+
+		expect(result.success).toBe(false)
+		expect(result.error).toMatch(/Invalid JSON arguments for searchMemories/)
+		expect(searchExecute).not.toHaveBeenCalled()
+	})
+
+	it("still passes well-formed arguments through", async () => {
+		searchExecute.mockResolvedValue({ results: [{ id: "mem_1" }] })
+		const execute = openAi.createToolCallExecutor(API_KEY, {
+			containerTags: ["user_1"],
+		})
+
+		const result = JSON.parse(
+			await execute(
+				toolCall(
+					"searchMemories",
+					JSON.stringify({ informationToGet: "tea", limit: 3 }),
+				),
+			),
+		)
+
+		expect(result.success).toBe(true)
+		expect(searchExecute).toHaveBeenCalledWith(
+			expect.objectContaining({ q: "tea", limit: 3 }),
+		)
+	})
+})
+
 describe("ClaudeMemoryTool", () => {
 	const FILE_PATH = "/memories/prefs.txt"
 	const CUSTOM_ID = "memories_prefs_txt"