
issue-implementation
by langadventurellc
Greatly improve how AI coding agents handle complex projects. Task Trellis helps track requirements for projects, breaks them down into smaller manageable parts until you have trackable and assignable tasks with built-in workflow management, dependency handling, and progress tracking. Basically, it's like Jira for coding agents.
SKILL.md
name: issue-implementation description: This skill should be used when the user asks to "implement task", "claim task", "work on task", "start working", or mentions implementing or claiming a Trellis task. Claims and implements a task following Research and Plan -> Implement workflow. allowed-tools:
- mcp__task-trellis__claim_task
- mcp__task-trellis__get_issue
- mcp__task-trellis__complete_task
- mcp__task-trellis__append_issue_log
- mcp__task-trellis__append_modified_files
- mcp__task-trellis__list_issues
- Task
- Glob
- Grep
- Read
- Edit
- Write
- Bash
- AskUserQuestion
Implement Task
Claim and implement the next available task from the backlog using the Research and Plan → Implement workflow.
Goal
Claim the highest-priority available task and implement it following project standards, with comprehensive research, planning, and quality checks before marking complete.
Process
1. Claim Next Available Task
Input
$ARGUMENTS (optional) - Can specify:
taskId- 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 withtaskId)
Use claim_task to claim. Tasks are managed in the .trellis folder.
2. Research and Planning Phase (MANDATORY)
Delegate research to planner, but verify critical details:
- Read parent issues for context: Use
get_issueto read the parent feature for context and requirements. Do not continue until you have claimed a task. - Generate Implementation Plan: Use the Implementation Plan Generator subagent with the task description, parent context, and project constraints.
- CRITICAL - If No Response From Subagent: STOP IMMEDIATELY and alert the user.
- CRITICAL - Verify the Plan: 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
Trust but Verify:
- The plan is your guide, not gospel—it may contain inaccuracies
- Question suspicious patterns; investigate before implementing
- Course-correct based on reality, not the plan
Common hallucination points: File paths, import statements, method names, config locations.
When You Find Discrepancies:
- 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 developer would ask more questions, not fewer. If you're making assumptions, stop and ask instead.
Ask questions when:
- Requirements are ambiguous or incomplete
- Multiple valid approaches exist
- You're unsure about architectural decisions
- The task scope seems unclear
4. Implementation Phase
Execute the plan with progress updates:
- Write clean code: Follow project conventions and best practices
- Implement incrementally: Build and test small pieces before moving on
- Run quality checks frequently: Format, lint, and test after each major change
- Write tests alongside code: Don't leave testing for the end
- Handle errors gracefully: Include proper error handling
5. Complete Task
Verify and document completion:
- Verify all requirements met: Check implementation satisfies task description
- Confirm quality checks pass: All tests, linting, and formatting clean
- Write meaningful summary: Describe what was implemented and key decisions
- List all changed files: Document what was created or modified
Use complete_task with task ID, summary, and files changed.
STOP! - Complete one task only. Do not implement another task.
Quality Standards
- Research First: Never skip research phase unless specifically instructed by the user
- Test Coverage: Write tests in same task
- Quality Checks: All tests must pass before marking task complete
Score
Total Score
Based on repository quality metrics
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
1ヶ月以内に更新
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
Reviews
Reviews coming soon


