skill-creator-plus（技能创建器加）

Skill Creator+

Design and review Agent Skills that trigger reliably and stay lean.

Decision Flow

Start by deciding whether the user really needs a skill.

If the answer is not Skill, stop and create the right primitive instead.

→ references/customization-primitives.md for the full selection guide

When to Use

• Create skill, /create-skill, new skill, review skill, fix skill trigger, SKILL.md, スキル作成
• Creating a new skill from scratch
• Updating or refactoring an existing skill
• Reviewing existing SKILL.md files
• Deciding whether a customization should be a skill before authoring it

Core Principles

Default assumption: Claude is already very smart. Challenge each piece: "Does this justify its token cost?"

Before adding a new section, ask three questions:

1. Can this update an existing rule instead?
2. Can this move to references/ instead of staying in SKILL.md?
3. Is this actually reusable, or is it just a session-specific note?

Skill Structure

→ references/skill-structure.md for locations, frontmatter, and bundled resource rules

skill-name/
├── SKILL.md (required)        # Lean overview + decision points
├── scripts/                   # Deterministic helpers
├── references/                # Load on demand
└── assets/                    # Templates and reusable outputs

→ See skill-structure.md > What NOT to Include for excluded files.

Creation Process

→ references/creation-process.md for the end-to-end workflow

Refactor Order

When improving an existing skill, use this order:

1. Delete stale or low-value guidance
2. Merge duplicate rules
3. Move long detail to references/
4. Add genuinely missing guidance last

Frontmatter and Triggering

Use the smallest viable frontmatter.

---
name: skill-name
description: "What it does. Use when [trigger conditions]. Triggers on 'keyword', 'phrase'."
argument-hint: "Optional slash-command hint"
user-invocable: true
disable-model-invocation: false
---

• name must match the folder name
• description must say both what and when
• Optional fields should be intentional, not boilerplate
• Resource links must stay relative and shallow

→ references/common-pitfalls.md for silent failures and trigger misses

SKILL.md Guidelines

Size Target

→ See skill-review-checklist.md > Line Count Target for size guidelines.

Quick rule: < 150 lines is good, > 300 lines must split to references.

When to Use Section

Start with generic keywords users are likely to say:

## When to Use

- **PDF**, **extract text**, **form filling** ← Keywords first
- Processing documents with embedded images
- Filling PDF forms programmatically

Body

• Use imperative/infinitive form
• Link to references for details
• Keep essential workflow only
• Add decision points when misuse is common
• Reuse a proven shape from references/skill-structure-gallery.md before inventing a custom layout
• Push reference material out of SKILL.md aggressively
• Prefer replacing or compressing existing text over appending a new subsection

Review Checklist

→ references/skill-review-checklist.md

For bloat review specifically, use references/skill-bloat-review.md.

- [ ] SKILL.md under 150 lines?
- [ ] Request truly needs a skill?
- [ ] Description has trigger conditions?
- [ ] Optional frontmatter fields are intentional?
- [ ] Details moved to references/?
- [ ] No README.md or auxiliary docs?

Key References

Done Criteria

• Request is confirmed to be a skill, not another primitive
• Scope is decided before file creation
• SKILL.md created and under 150 lines
• Frontmatter has name + description with trigger conditions
• Optional fields are added only when they change behavior
• Details moved to references/ (Progressive Disclosure)
• Review checklist passed