From 76c9839bc94b2da37cdbac6e763a9aabb84a7467 Mon Sep 17 00:00:00 2001 From: zbigniew sobiecki Date: Sun, 18 Jan 2026 10:59:08 +0100 Subject: [PATCH] feat(gadgets): split TodoUpsert into focused gadgets Extract status updates from TodoUpsert into a dedicated TodoUpdateStatus gadget to reduce LLM confusion and improve task tracking clarity. Changes: - Add TodoUpdateStatus gadget for status-only updates (pending/in_progress/done) - Simplify TodoUpsert to handle content creation/updates only - Update prompts to guide agents on using the separate gadgets - Add git status and PR status to implementation trailing messages - Add EditFile/WriteFile result handling guidance to prompts - Add comprehensive tests for both gadgets The separation makes each gadget's purpose clearer: - TodoUpsert: Plan work by creating todos - TodoUpdateStatus: Track progress by updating status Co-Authored-By: Claude --- src/agents/base.ts | 3 +- .../prompts/templates/implementation.eta | 15 +- .../templates/partials/rules-efficiency.eta | 42 +++ .../prompts/templates/respond-to-review.eta | 15 +- src/agents/review.ts | 3 +- src/agents/utils/agentLoop.ts | 6 +- src/config/hintConfig.ts | 52 +++- src/gadgets/todo/TodoUpdateStatus.ts | 46 +++ src/gadgets/todo/TodoUpsert.ts | 86 ++---- src/gadgets/todo/index.ts | 1 + tests/unit/gadgets/todo.test.ts | 284 ++++++++++++++++++ 11 files changed, 473 insertions(+), 80 deletions(-) create mode 100644 src/gadgets/todo/TodoUpdateStatus.ts create mode 100644 tests/unit/gadgets/todo.test.ts diff --git a/src/agents/base.ts b/src/agents/base.ts index dc8bad31..0d344c5a 100644 --- a/src/agents/base.ts +++ b/src/agents/base.ts @@ -15,7 +15,7 @@ import { ReadFile } from '../gadgets/ReadFile.js'; import { Sleep } from '../gadgets/Sleep.js'; import { CreatePR } from '../gadgets/github/index.js'; import { Tmux } from '../gadgets/tmux.js'; -import { TodoDelete, TodoUpsert } from '../gadgets/todo/index.js'; +import { TodoDelete, TodoUpdateStatus, TodoUpsert } from '../gadgets/todo/index.js'; import { AddChecklistToCard, CreateTrelloCard, @@ -290,6 +290,7 @@ function createAgentBuilderWithGadgets( new Sleep(), // Task tracking gadgets new TodoUpsert(), + new TodoUpdateStatus(), new TodoDelete(), // GitHub gadgets (no PR creation for planning) ...(isReadOnlyAgent ? [] : [new CreatePR()]), diff --git a/src/agents/prompts/templates/implementation.eta b/src/agents/prompts/templates/implementation.eta index 0454a767..31683354 100644 --- a/src/agents/prompts/templates/implementation.eta +++ b/src/agents/prompts/templates/implementation.eta @@ -21,15 +21,12 @@ You are an expert software engineer implementing features and fixing issues base ## Task Tracking -Use TodoUpsert to create a todo list after reading the Trello card. This helps you: -- Plan your work upfront -- Track progress through implementation -- Stay focused and avoid getting lost in exploration - -**Workflow:** -1. After reading the card, create todos for each major step (e.g., "Create feature branch", "Implement X", "Write tests", "Run lint/type checks", "Create PR") -2. Mark each todo as `in_progress` when you start it -3. Mark each todo as `done` when you complete it +Use the todo gadgets to plan and track your progress: + +1. **Plan**: After reading the Trello card, use `TodoUpsert` to create todos for major steps +2. **Start**: Use `TodoUpdateStatus` to mark a todo as `in_progress` when you begin working on it +3. **Complete**: Use `TodoUpdateStatus` to mark it as `done` when finished +4. **One at a time**: Only have ONE todo in_progress at any time ## Understanding the Codebase diff --git a/src/agents/prompts/templates/partials/rules-efficiency.eta b/src/agents/prompts/templates/partials/rules-efficiency.eta index 1c2ab5e7..290c834e 100644 --- a/src/agents/prompts/templates/partials/rules-efficiency.eta +++ b/src/agents/prompts/templates/partials/rules-efficiency.eta @@ -54,3 +54,45 @@ in a single batch. Avoid the 'fix-run-repeat' loop for known breaking changes. When validating changes, always prioritize auto-fixing commands (e.g., format, lint --fix) before running read-only checks. If a generated artifact (like a migration or lockfile) looks incorrect, investigate the environment state rather than manually patching the output. + +### Acting on EditFile/WriteFile Results (CRITICAL) + +EditFile and WriteFile return structured output. You MUST read and act on it. + +**Status Codes:** +- `status=success` - Edit worked, but CHECK THE DIAGNOSTICS SECTION +- `status=failed` - Search content not found, USE THE SUGGESTIONS +- `status=error` - Operation failed (permissions, path), read error message + +**On status=failed (search not found):** + +The output includes a SUGGESTIONS section with similar content found in the file: +``` +SUGGESTIONS (similar content found): +Line 42 (85% similar): +```{what actually exists}``` +``` + +1. READ this section - it shows what the file actually contains +2. Adjust your search pattern to match the actual content +3. Common issues: whitespace differences, indentation, content changed by previous edit +4. NEVER retry the exact same search - that's a loop + +**On status=success (check diagnostics):** + +Success responses include TypeScript and Biome diagnostics: +``` +=== TypeScript Check === +{any type errors from your edit} + +=== Biome Lint === +{any lint issues from your edit} +``` + +If diagnostics show issues, FIX THEM IMMEDIATELY before making more edits. +Don't proceed to other files until the current file is clean. + +**Recovery Escalation:** +1. First failure: Use SUGGESTIONS to adjust search pattern +2. Second failure: Read entire file with ReadFile, understand actual structure +3. Third failure: Use WriteFile to replace the entire file content \ No newline at end of file diff --git a/src/agents/prompts/templates/respond-to-review.eta b/src/agents/prompts/templates/respond-to-review.eta index a881e7b7..5af6df61 100644 --- a/src/agents/prompts/templates/respond-to-review.eta +++ b/src/agents/prompts/templates/respond-to-review.eta @@ -13,15 +13,12 @@ You are an expert software engineer addressing code review feedback on a pull re ## Task Tracking -Use TodoUpsert to create a todo list after reading the review comments. This helps you: -- Track each review comment that needs addressing -- Avoid missing any feedback -- Stay focused and organized - -**Workflow:** -1. After reading comments, create a todo for each comment/issue to address -2. Mark each todo as `in_progress` when you start it -3. Mark each todo as `done` when you complete it AND reply to the comment +Use the todo gadgets to track review comments: + +1. **Plan**: After reading comments, use `TodoUpsert` to create a todo for each comment to address +2. **Start**: Use `TodoUpdateStatus` to mark a todo as `in_progress` when you begin working on it +3. **Complete**: Use `TodoUpdateStatus` to mark it as `done` when finished AND reply to the comment +4. **One at a time**: Only have ONE todo in_progress at any time ## Understanding the Codebase diff --git a/src/agents/review.ts b/src/agents/review.ts index 6f7ea810..74def9ad 100644 --- a/src/agents/review.ts +++ b/src/agents/review.ts @@ -20,7 +20,7 @@ import { formatCheckStatus, } from '../gadgets/github/index.js'; import { Tmux } from '../gadgets/tmux.js'; -import { TodoDelete, TodoUpsert } from '../gadgets/todo/index.js'; +import { TodoDelete, TodoUpdateStatus, TodoUpsert } from '../gadgets/todo/index.js'; import { githubClient } from '../github/client.js'; import type { AgentInput, AgentResult, CascadeConfig, ProjectConfig } from '../types/index.js'; import { cleanupLogDirectory, cleanupLogFile, createFileLogger } from '../utils/fileLogger.js'; @@ -253,6 +253,7 @@ function createReviewAgentBuilder( new Sleep(), // Task tracking gadgets new TodoUpsert(), + new TodoUpdateStatus(), new TodoDelete(), // GitHub gadgets (read + create review) new GetPRDetails(), diff --git a/src/agents/utils/agentLoop.ts b/src/agents/utils/agentLoop.ts index 9b897a8c..a3b6916d 100644 --- a/src/agents/utils/agentLoop.ts +++ b/src/agents/utils/agentLoop.ts @@ -90,9 +90,13 @@ function addGadgetSpecificLogContext( if (gadgetName === 'TodoUpsert') { if (parameters.id) logContext.id = parameters.id; - if (parameters.status) logContext.status = parameters.status; if (parameters.content) logContext.todo = truncateContent(String(parameters.content), 80); } + + if (gadgetName === 'TodoUpdateStatus') { + logContext.id = parameters.id; + logContext.status = parameters.status; + } } // ============================================================================ diff --git a/src/config/hintConfig.ts b/src/config/hintConfig.ts index e26aa7a6..4204059c 100644 --- a/src/config/hintConfig.ts +++ b/src/config/hintConfig.ts @@ -1,3 +1,4 @@ +import { execSync } from 'node:child_process'; import type { TrailingMessage } from 'llmist'; import { formatTodoList, loadTodos } from '../gadgets/todo/storage.js'; @@ -33,6 +34,31 @@ function getAgentHint(agentType?: string): string { return AGENT_HINTS.default; } +/** + * Run a shell command and return output, or null on error. + */ +function runCommand(command: string): string | null { + try { + return execSync(command, { encoding: 'utf-8', timeout: 5000 }).trim(); + } catch { + return null; + } +} + +/** + * Get git status output (short format for brevity). + */ +function getGitStatus(): string | null { + return runCommand('git status --short'); +} + +/** + * Get PR view output if a PR exists for current branch. + */ +function getPRView(): string | null { + return runCommand('gh pr view 2>/dev/null'); +} + /** * Format the iteration status line with appropriate urgency indicator. */ @@ -76,13 +102,33 @@ export function getIterationTrailingMessage(agentType?: string): TrailingMessage return (ctx) => { const iterationStatus = formatIterationStatus(ctx.iteration, ctx.maxIterations, batchHint); - // For implementation agent, include the current todo list + // For implementation agent, include progress info, git status, and PR status if (agentType === 'implementation') { + const sections: string[] = [iterationStatus]; + + // Add todo list if there are todos const todos = loadTodos(); if (todos.length > 0) { - const todoListFormatted = formatTodoList(todos); - return `${iterationStatus}\n\n## Current Progress\n\n${todoListFormatted}`; + sections.push(`## Current Progress\n\n${formatTodoList(todos)}`); } + + // Add git status + const gitStatus = getGitStatus(); + if (gitStatus) { + sections.push(`## Git Status\n\n\`\`\`\n${gitStatus}\n\`\`\``); + } else { + sections.push('## Git Status\n\nNo uncommitted changes.'); + } + + // Add PR status if a PR exists + const prView = getPRView(); + if (prView) { + sections.push(`## PR Status\n\n\`\`\`\n${prView}\n\`\`\``); + } else { + sections.push('## PR Status\n\nNo PR exists for current branch.'); + } + + return sections.join('\n\n'); } return iterationStatus; diff --git a/src/gadgets/todo/TodoUpdateStatus.ts b/src/gadgets/todo/TodoUpdateStatus.ts new file mode 100644 index 00000000..db407b8e --- /dev/null +++ b/src/gadgets/todo/TodoUpdateStatus.ts @@ -0,0 +1,46 @@ +/** + * TodoUpdateStatus gadget - Update the status of existing todo items. + * Helps agents track progress through implementation tasks. + */ +import { Gadget, z } from 'llmist'; +import { type TodoStatus, formatTodoList, loadTodos, saveTodos } from './storage.js'; + +export class TodoUpdateStatus extends Gadget({ + name: 'TodoUpdateStatus', + description: `Update the status of an existing todo item. +Use this to track progress: mark todos as in_progress when starting, done when complete.`, + schema: z.object({ + id: z.string().describe('ID of the todo to update (required)'), + status: z.enum(['pending', 'in_progress', 'done']).describe('New status'), + comment: z.string().optional().describe('Brief explanation of why this change is needed'), + }), + examples: [ + { + params: { id: '1', status: 'in_progress', comment: 'Starting work on first task' }, + output: + 'āœļø Updated todo #1 ā†’ in_progress.\n\nšŸ“‹ Todo List\n Progress: 0/1 done, 1 in progress, 0 pending\n\nšŸ”„ #1 [in_progress]: Read and understand requirements', + comment: 'Mark todo as in progress', + }, + { + params: { id: '1', status: 'done', comment: 'Task completed successfully' }, + output: + 'āœļø Updated todo #1 ā†’ done.\n\nšŸ“‹ Todo List\n Progress: 1/1 done, 0 in progress, 0 pending\n\nāœ… #1 [done]: Read and understand requirements', + comment: 'Mark todo as done', + }, + ], +}) { + override execute(params: this['params']): string { + const { id, status } = params; + const todos = loadTodos(); + const index = todos.findIndex((t) => t.id === id); + + if (index === -1) { + throw new Error(`Todo #${id} not found. Use TodoUpsert to create todos first.`); + } + + todos[index].status = status as TodoStatus; + todos[index].updatedAt = new Date().toISOString(); + saveTodos(todos); + return `āœļø Updated todo #${id} ā†’ ${status}.\n\n${formatTodoList(todos)}`; + } +} diff --git a/src/gadgets/todo/TodoUpsert.ts b/src/gadgets/todo/TodoUpsert.ts index 45a4b04b..d9b3fe19 100644 --- a/src/gadgets/todo/TodoUpsert.ts +++ b/src/gadgets/todo/TodoUpsert.ts @@ -1,21 +1,13 @@ /** - * TodoUpsert gadget - Create or update todo items. - * Helps agents track their progress through implementation tasks. + * TodoUpsert gadget - Create or update todo item content. + * Helps agents plan and organize their implementation tasks. */ import { Gadget, z } from 'llmist'; -import { - type Todo, - type TodoStatus, - formatTodoList, - getNextId, - loadTodos, - saveTodos, -} from './storage.js'; +import { type Todo, formatTodoList, getNextId, loadTodos, saveTodos } from './storage.js'; interface TodoItem { id?: string; content?: string; - status?: string; } interface BatchResult { @@ -31,14 +23,13 @@ function upsertBatchItem(item: TodoItem, todos: Todo[], now: string): 'created' todos.push({ id: item.id, content: item.content, - status: (item.status as TodoStatus) ?? 'pending', + status: 'pending', createdAt: now, updatedAt: now, }); return 'created'; } if (item.content !== undefined) todos[index].content = item.content; - if (item.status !== undefined) todos[index].status = item.status as TodoStatus; todos[index].updatedAt = now; return 'updated'; } @@ -47,7 +38,7 @@ function upsertBatchItem(item: TodoItem, todos: Todo[], now: string): 'created' todos.push({ id: getNextId(todos), content: item.content, - status: (item.status as TodoStatus) ?? 'pending', + status: 'pending', createdAt: now, updatedAt: now, }); @@ -76,18 +67,6 @@ function formatBatchResult(result: BatchResult, todos: Todo[]): string { return `${parts.join(', ')}.\n\n${formatTodoList(todos)}`; } -function updateExistingTodo( - index: number, - todos: Todo[], - content: string | undefined, - status: string | undefined, - now: string, -): void { - if (content !== undefined) todos[index].content = content; - if (status !== undefined) todos[index].status = status as TodoStatus; - todos[index].updatedAt = now; -} - const todoItemSchema = z.object({ id: z.string().optional().describe('ID of existing todo to update. Omit to create a new todo.'), content: z @@ -95,20 +74,17 @@ const todoItemSchema = z.object({ .min(1) .optional() .describe('The todo item description. Required when creating, optional when updating.'), - status: z - .enum(['pending', 'in_progress', 'done']) - .optional() - .describe("Todo status: pending, in_progress, or done. Defaults to 'pending' for new items."), }); export class TodoUpsert extends Gadget({ name: 'TodoUpsert', - description: `Create or update one or more todo items. + description: `Create or update todo item content. -Use this to plan your work at the start of a task and track progress as you go. -- For a single item: use id/content/status directly +Use this to plan your work at the start of a task. +- For a single item: use id/content directly - For multiple items: use the 'items' array to batch create/update +All new todos start with status 'pending'. Use TodoUpdateStatus to change status. Returns the full todo list after the operation.`, schema: z.object({ id: z.string().optional().describe('ID of existing todo to update. Omit to create a new todo.'), @@ -117,38 +93,24 @@ Returns the full todo list after the operation.`, .min(1) .optional() .describe('The todo item description. Required when creating, optional when updating.'), - status: z - .enum(['pending', 'in_progress', 'done']) - .optional() - .describe("Todo status: pending, in_progress, or done. Defaults to 'pending' for new items."), items: z .array(todoItemSchema) .optional() .describe( - 'Batch mode: array of todo items to create/update. Each item has id/content/status. Use this to create multiple todos at once.', + 'Batch mode: array of todo items to create/update. Each item has id/content. Use this to create multiple todos at once.', ), + comment: z.string().optional().describe('Brief explanation of why this change is needed'), }), examples: [ { params: { content: 'Read and understand the Trello card requirements', + comment: 'Planning initial task', }, output: 'āž• Created todo #1.\n\nšŸ“‹ Todo List\n Progress: 0/1 done, 0 in progress, 1 pending\n\nā¬œ #1 [pending]: Read and understand the Trello card requirements', comment: 'Create a new todo item', }, - { - params: { id: '1', status: 'in_progress' }, - output: - 'āœļø Updated todo #1.\n\nšŸ“‹ Todo List\n Progress: 0/1 done, 1 in progress, 0 pending\n\nšŸ”„ #1 [in_progress]: Read and understand the Trello card requirements', - comment: 'Mark a todo as in progress', - }, - { - params: { id: '1', status: 'done' }, - output: - 'āœļø Updated todo #1.\n\nšŸ“‹ Todo List\n Progress: 1/1 done, 0 in progress, 0 pending\n\nāœ… #1 [done]: Read and understand the Trello card requirements', - comment: 'Mark a todo as done', - }, { params: { items: [ @@ -157,6 +119,7 @@ Returns the full todo list after the operation.`, { content: 'Write tests' }, { content: 'Run lint and typecheck' }, ], + comment: 'Planning implementation steps', }, output: 'āž• Created 4 todos.\n\nšŸ“‹ Todo List\n Progress: 0/4 done, 0 in progress, 4 pending\n\nā¬œ #1 [pending]: Create feature branch\nā¬œ #2 [pending]: Implement feature\nā¬œ #3 [pending]: Write tests\nā¬œ #4 [pending]: Run lint and typecheck', @@ -165,7 +128,17 @@ Returns the full todo list after the operation.`, ], }) { override execute(params: this['params']): string { - const { id, content, status, items } = params; + const { id, content, items } = params; + + // Prevent mixing single-item update with batch create (silently fails otherwise) + if (items && items.length > 0 && (id !== undefined || content !== undefined)) { + throw new Error( + "Cannot combine top-level id/content with 'items' array. " + + 'Use either single-item mode (id/content) OR batch mode (items array), not both. ' + + 'To update one todo and create others, make two separate calls.', + ); + } + const todos = loadTodos(); const now = new Date().toISOString(); @@ -181,12 +154,12 @@ Returns the full todo list after the operation.`, const index = todos.findIndex((t) => t.id === id); if (index === -1) { if (!content) { - return `āŒ Error: 'content' is required when creating a new todo.`; + throw new Error("'content' is required when creating a new todo."); } todos.push({ id, content, - status: (status as TodoStatus) ?? 'pending', + status: 'pending', createdAt: now, updatedAt: now, }); @@ -194,21 +167,22 @@ Returns the full todo list after the operation.`, return `āž• Created todo #${id}.\n\n${formatTodoList(todos)}`; } - updateExistingTodo(index, todos, content, status, now); + if (content !== undefined) todos[index].content = content; + todos[index].updatedAt = now; saveTodos(todos); return `āœļø Updated todo #${id}.\n\n${formatTodoList(todos)}`; } // Create new todo - content is required if (!content) { - return `āŒ Error: 'content' is required when creating a new todo.`; + throw new Error("'content' is required when creating a new todo."); } const newId = getNextId(todos); todos.push({ id: newId, content, - status: (status as TodoStatus) ?? 'pending', + status: 'pending', createdAt: now, updatedAt: now, }); diff --git a/src/gadgets/todo/index.ts b/src/gadgets/todo/index.ts index 0cb5c330..55cc36a3 100644 --- a/src/gadgets/todo/index.ts +++ b/src/gadgets/todo/index.ts @@ -2,6 +2,7 @@ * Todo gadgets for agent task tracking. * Helps agents plan work and track progress through implementation tasks. */ +export { TodoUpdateStatus } from './TodoUpdateStatus.js'; export { TodoUpsert } from './TodoUpsert.js'; export { TodoDelete } from './TodoDelete.js'; export { initTodoSession } from './storage.js'; diff --git a/tests/unit/gadgets/todo.test.ts b/tests/unit/gadgets/todo.test.ts new file mode 100644 index 00000000..6757bfe2 --- /dev/null +++ b/tests/unit/gadgets/todo.test.ts @@ -0,0 +1,284 @@ +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'; + +import { TodoUpdateStatus } from '../../../src/gadgets/todo/TodoUpdateStatus.js'; +import { TodoUpsert } from '../../../src/gadgets/todo/TodoUpsert.js'; + +// Mock the storage module +vi.mock('../../../src/gadgets/todo/storage.js', () => { + let todos: Array<{ + id: string; + content: string; + status: string; + createdAt: string; + updatedAt: string; + }> = []; + + return { + loadTodos: vi.fn(() => todos), + saveTodos: vi.fn((newTodos) => { + todos = newTodos; + }), + getNextId: vi.fn((existingTodos) => { + const maxId = existingTodos.reduce( + (max: number, t: { id: string }) => Math.max(max, Number.parseInt(t.id) || 0), + 0, + ); + return String(maxId + 1); + }), + formatTodoList: vi.fn((existingTodos) => { + if (existingTodos.length === 0) return 'šŸ“‹ Todo list is empty.'; + const lines = existingTodos.map( + (t: { id: string; status: string; content: string }) => + `#${t.id} [${t.status}]: ${t.content}`, + ); + return `šŸ“‹ Todo List\n${lines.join('\n')}`; + }), + // Helper to reset todos between tests + _resetTodos: () => { + todos = []; + }, + _setTodos: ( + newTodos: Array<{ + id: string; + content: string; + status: string; + createdAt: string; + updatedAt: string; + }>, + ) => { + todos = newTodos; + }, + }; +}); + +// Import the mocked functions to access helpers +import * as storage from '../../../src/gadgets/todo/storage.js'; + +describe('TodoUpsert', () => { + let gadget: TodoUpsert; + + beforeEach(() => { + gadget = new TodoUpsert(); + // Reset todos before each test + (storage as unknown as { _resetTodos: () => void })._resetTodos(); + }); + + afterEach(() => { + vi.clearAllMocks(); + }); + + describe('gadget metadata', () => { + it('has correct name', () => { + expect(gadget.name).toBe('TodoUpsert'); + }); + + it('has description mentioning create or update', () => { + expect(gadget.description).toContain('Create or update'); + }); + + it('description mentions TodoUpdateStatus for status changes', () => { + expect(gadget.description).toContain('TodoUpdateStatus'); + }); + }); + + describe('mixed mode validation', () => { + it('throws error when combining id with items array', () => { + expect(() => + gadget.execute({ + id: '1', + items: [{ content: 'New task' }], + }), + ).toThrow("Cannot combine top-level id/content with 'items' array"); + }); + + it('throws error when combining content with items array', () => { + expect(() => + gadget.execute({ + content: 'Some content', + items: [{ content: 'New task' }], + }), + ).toThrow("Cannot combine top-level id/content with 'items' array"); + }); + + it('error message suggests making two separate calls', () => { + try { + gadget.execute({ + id: '1', + items: [{ content: 'New task' }], + }); + expect.fail('Should have thrown'); + } catch (error) { + expect((error as Error).message).toContain('make two separate calls'); + } + }); + + it('allows empty items array with id (single-item mode)', () => { + // Set up an existing todo + (storage as unknown as { _setTodos: (todos: unknown[]) => void })._setTodos([ + { id: '1', content: 'Existing task', status: 'pending', createdAt: '', updatedAt: '' }, + ]); + + // Empty items array should be ignored, single-item mode takes over + const result = gadget.execute({ + id: '1', + content: 'Updated task', + items: [], + }); + + expect(result).toContain('Updated todo #1'); + }); + }); + + describe('single-item mode', () => { + it('creates a new todo with content', () => { + const result = gadget.execute({ + content: 'New task', + }); + + expect(result).toContain('Created todo #1'); + expect(storage.saveTodos).toHaveBeenCalled(); + }); + + it('new todos default to pending status', () => { + gadget.execute({ + content: 'New task', + }); + + const savedTodos = (storage.saveTodos as ReturnType).mock.calls[0][0]; + expect(savedTodos[0].status).toBe('pending'); + }); + + it('updates existing todo content', () => { + (storage as unknown as { _setTodos: (todos: unknown[]) => void })._setTodos([ + { id: '1', content: 'Existing task', status: 'pending', createdAt: '', updatedAt: '' }, + ]); + + const result = gadget.execute({ + id: '1', + content: 'Updated task', + }); + + expect(result).toContain('Updated todo #1'); + }); + + it('throws error when creating without content', () => { + expect(() => gadget.execute({})).toThrow("'content' is required when creating a new todo."); + }); + + it('throws error when creating with id but no content', () => { + expect(() => gadget.execute({ id: '99' })).toThrow( + "'content' is required when creating a new todo.", + ); + }); + }); + + describe('batch mode', () => { + it('creates multiple todos at once', () => { + const result = gadget.execute({ + items: [{ content: 'Task 1' }, { content: 'Task 2' }, { content: 'Task 3' }], + }); + + expect(result).toContain('Created 3 todos'); + expect(storage.saveTodos).toHaveBeenCalled(); + }); + + it('batch created todos default to pending status', () => { + gadget.execute({ + items: [{ content: 'Task 1' }, { content: 'Task 2' }], + }); + + const savedTodos = (storage.saveTodos as ReturnType).mock.calls[0][0]; + expect(savedTodos[0].status).toBe('pending'); + expect(savedTodos[1].status).toBe('pending'); + }); + }); +}); + +describe('TodoUpdateStatus', () => { + let gadget: TodoUpdateStatus; + + beforeEach(() => { + gadget = new TodoUpdateStatus(); + // Reset todos before each test + (storage as unknown as { _resetTodos: () => void })._resetTodos(); + }); + + afterEach(() => { + vi.clearAllMocks(); + }); + + describe('gadget metadata', () => { + it('has correct name', () => { + expect(gadget.name).toBe('TodoUpdateStatus'); + }); + + it('has description mentioning status update', () => { + expect(gadget.description).toContain('status'); + }); + }); + + describe('status updates', () => { + it('updates status of existing todo to in_progress', () => { + (storage as unknown as { _setTodos: (todos: unknown[]) => void })._setTodos([ + { id: '1', content: 'Existing task', status: 'pending', createdAt: '', updatedAt: '' }, + ]); + + const result = gadget.execute({ + id: '1', + status: 'in_progress', + }); + + expect(result).toContain('Updated todo #1'); + expect(result).toContain('in_progress'); + expect(storage.saveTodos).toHaveBeenCalled(); + }); + + it('updates status of existing todo to done', () => { + (storage as unknown as { _setTodos: (todos: unknown[]) => void })._setTodos([ + { id: '1', content: 'Existing task', status: 'in_progress', createdAt: '', updatedAt: '' }, + ]); + + const result = gadget.execute({ + id: '1', + status: 'done', + }); + + expect(result).toContain('Updated todo #1'); + expect(result).toContain('done'); + }); + + it('updates updatedAt timestamp', () => { + (storage as unknown as { _setTodos: (todos: unknown[]) => void })._setTodos([ + { id: '1', content: 'Existing task', status: 'pending', createdAt: '', updatedAt: '' }, + ]); + + gadget.execute({ + id: '1', + status: 'done', + }); + + const savedTodos = (storage.saveTodos as ReturnType).mock.calls[0][0]; + expect(savedTodos[0].updatedAt).not.toBe(''); + }); + }); + + describe('error handling', () => { + it('throws error for non-existent todo', () => { + expect(() => + gadget.execute({ + id: '99', + status: 'done', + }), + ).toThrow('Todo #99 not found. Use TodoUpsert to create todos first.'); + }); + + it('throws error when todo list is empty', () => { + expect(() => + gadget.execute({ + id: '1', + status: 'in_progress', + }), + ).toThrow('Todo #1 not found'); + }); + }); +});