Skip to content
Back to skills

Issue Creation

ASecurity

This skill should be used when the user asks to create Trellis issues including "create a project", "create epics", "create features", "create tasks", "new project", "new epic", "new feature", "new task", "break down project into epics", "break down epic into features", "break down feature into tasks", "decompose project", "decompose epic", "decompose feature", or mentions creating any type of issue in Trellis.

  • 666 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added February 9, 2026
business

Works with

  • mcp

Security analysis

A100/100

Scanned February 12, 2026

npx -y skills add majiayu000/claude-skill-registry --skill issue-creation --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Issue Creation?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Issue Creation
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/majiayu000-issue-creation/badge)](https://www.skillsdirectory.com/skills/majiayu000-issue-creation)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: issue-creation
description: This skill should be used when the user asks to create Trellis issues including "create a project", "create epics", "create features", "create tasks", "new project", "new epic", "new feature", "new task", "break down project into epics", "break down epic into features", "break down feature into tasks", "decompose project", "decompose epic", "decompose feature", or mentions creating any type of issue in Trellis.
allowed-tools:
  - mcp__task-trellis__create_issue
  - mcp__task-trellis__get_issue
  - mcp__task-trellis__update_issue
  - mcp__task-trellis__list_issues
  - mcp__perplexity-ask__perplexity_ask
  - Task
  - Glob
  - Grep
  - Read
  - AskUserQuestion
---

# Create Trellis Issues

Create issues in the Trellis task management system. This skill supports creating all issue types: projects, epics, features, and tasks.

## Issue Type Hierarchy

Trellis uses a hierarchical issue structure:

```
Project -> Epic -> Feature -> Task
```

- **Project**: Top-level container representing a complete initiative
- **Epic**: Major work stream within a project
- **Feature**: Implementable functionality within an epic
- **Task**: Atomic unit of work (1-2 hours) within a feature

Each type can be created standalone or within its parent hierarchy.

## Determining Issue Type

Based on the user's request, determine which issue type to create:

| User Request                                                         | Issue Type | Reference                |
| -------------------------------------------------------------------- | ---------- | ------------------------ |
| "create a project", "new project", "set up project management"       | Project    | [project.md](project.md) |
| "create epics", "break down project into epics", "decompose project" | Epic       | [epic.md](epic.md)       |
| "create features", "break down epic into features", "decompose epic" | Feature    | [feature.md](feature.md) |
| "create tasks", "break down feature into tasks", "decompose feature" | Task       | [task.md](task.md)       |

## Instructions

1. **Identify the issue type** the user wants to create based on their request
2. **Read the appropriate type-specific file** for detailed creation instructions:
   - For projects: Read [project.md](project.md)
   - For epics: Read [epic.md](epic.md)
   - For features: Read [feature.md](feature.md)
   - For tasks: Read [task.md](task.md)
3. **Follow the detailed process** in that file to gather requirements and create the issue(s)

## Common Principles

All issue types share these principles:

- **Research the codebase FIRST** - Before creating any issues, search the codebase to understand current state. Parent issues may be outdated. The codebase is the source of truth.
- **Proceed autonomously with parent IDs** - When given a parent issue ID (P-, E-, F-), create child issues without asking for confirmation.
- **Default to coarser granularity** - Prefer fewer, larger issues that are easier for AI agents to orchestrate. Don't create many tiny issues.
- **Ask questions only when necessary** - Only ask when requirements are genuinely ambiguous, critical information is missing, or decisions have significant irreversible consequences.
- **Include acceptance criteria** - All issues should have measurable success criteria.
- **Keep it simple** - Follow KISS, YAGNI principles. Don't over-engineer the structure.
- **Create sequentially** - When creating multiple issues, do them one at a time, not in parallel.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.