---
name: task-executor
description: Implements feature tasks by loading required skills, reading PRD/TechSpec context, analyzing dependencies, and executing the implementation with tests. Marks tasks as complete in tasks.md and triggers the task-reviewer agent upon completion. Use when the user asks to implement a task, execute a task, or start working on a specific task number. Do not use for creating tasks, running QA, code review, or bug fixing.
---

# Task Execution

## Procedures

**Step 1: Pre-Task Configuration (Mandatory)**

1. Read the task definition file at `./tasks/prd-[feature-slug]/[num]_task.md`.
2. Read the PRD at `./tasks/prd-[feature-slug]/prd.md` for context.
3. Read the Tech Spec at `./tasks/prd-[feature-slug]/techspec.md` for technical requirements.
4. Identify dependencies from previous tasks and verify they are complete.
5. Do NOT skip any of these reads.

**Step 2: Load Required Skills**

1. Identify the technologies involved in the task (React, Hono, shadcn, etc.).
2. Load the corresponding skills from `.claude/skills/` based on technologies used.
3. Use Context7 MCP to analyze documentation of involved languages, frameworks, and libraries.

**Step 3: Task Analysis (Mandatory)**

1. Analyze the task considering:
   - Main objectives.
   - How the task fits into the project context.
   - Alignment with project rules and standards.
   - Possible approaches or solutions.
2. Generate a task summary:
   - Task ID and Name.
   - PRD Context (main points).
   - Tech Spec Requirements (key technical requirements).
   - Dependencies.
   - Main Objectives.
   - Risks/Challenges.

**Step 4: Approach Plan (Mandatory)**

1. Define a numbered step-by-step approach.
2. Do NOT skip any step.

**Step 5: Implementation (Mandatory)**

1. Begin implementation immediately after planning.
2. Follow all project standards established in CLAUDE.md and project rules.
3. Implement solutions without workarounds.
4. Create and run all task tests before considering the task finished.

**Step 6: Mark Task Complete (Mandatory)**

1. After successful implementation and tests, mark the task as complete in `tasks.md`.

**Step 7: Review (Mandatory)**

1. Execute the `task-reviewer` agent to review the implementation.
2. Address any issues identified by the reviewer.
3. Do not finalize the task until review issues are resolved.

## Error Handling

- If the task file does not exist, halt and report to the user.
- If dependencies are not complete, warn the user and ask whether to proceed.
- If tests fail, fix the issues before marking the task as complete.
- If the task-reviewer identifies critical issues, address them before finalizing.