issue-implementation

# Implement Trellis Task

Claim and implement a single task from the Trellis task management system using the Research and Plan → Implement workflow.

**Note**: For implementing features, epics, or projects (which orchestrate multiple tasks), use the `issue-implementation-orchestration` skill instead.

## Input

`$ARGUMENTS` (optional) - Can specify:

- **Task ID**: Specific task ID to claim (e.g., "T-create-user-model")
- **Scope**: Hierarchical scope for task filtering (P-, E-, F- prefixed)
- **Force**: Bypass validation when claiming specific task (only with task ID)

**If no task ID specified**: Claims the next available task based on priority and readiness (prerequisites satisfied).

## Process

### 1. Claim Task

Use `claim_task` to claim the task. Tasks are managed in the `.trellis` folder.

### 2. Research and Planning Phase (MANDATORY)

**Research the codebase and plan your approach:**

- **Read parent issues for context**: Use `get_issue` to read the parent feature for context and requirements. Do not continue until you have claimed a task.
- **Research codebase patterns**: Search for similar implementations, conventions, and patterns in the codebase
- **Plan your approach**: Identify the files to modify, patterns to follow, and dependencies needed
- **CRITICAL - Verify your findings**: Spot-check before implementing:
  - Verify 2-3 key file paths actually exist
  - Confirm at least one pattern/convention identified
  - Check that referenced imports or dependencies are real

**When You Find Issues:**

- **Minor issues** (wrong path, naming): Adapt and continue
- **Major issues** (approach wrong, files don't exist): **STOP** and alert the user
- **Pattern mismatches**: Follow actual codebase patterns
- **Missing dependencies**: Check if installation needed or find alternatives

### 3. Clarify Before Implementing

**When in doubt, ask.** Use AskUserQuestion to clarify requirements or approach. Agents tend to be overconfident about what they can infer—a human