메뉴
Guides you through designing and writing a new Claude skill from scratch — determines the right complexity tier, writes the SKILL.md with correct frontmatter and instructions, and scaffolds any supporting assets, references, or scripts the skill needs.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Guides you through designing and packaging a complete skillset — determines what skills belong together, writes the SKILLSET.md manifest, identifies shared assets, and produces the full directory layout with all skill files ready to install.
Generates structured competitor breakdowns from information you provide. Uses a bundled analysis framework to evaluate competitors across consistent dimensions — positioning, pricing, features, weaknesses, and moat — and outputs a formatted report ready to share with stakeholders or paste into a product document.
Produces structured summaries of academic or technical papers. Fills a consistent template covering abstract, key contributions, method, results, limitations, and relevance. Calibrates terminology and depth to a named audience using the shared audience personas reference — so the summary is useful to the right reader, not just accurate.
Explains complex technical topics in plain language calibrated to a named audience. Loads a shared audience personas reference to determine vocabulary level, assumed prior knowledge, and preferred explanation style — so the explanation lands for that specific reader, not for a generic one.
Produces CHANGELOG entries from git history following the Keep a Changelog format. Uses a bundled script to extract commits since the last tag, classifies them into changelog sections using the shared commit conventions reference, and outputs a ready-to-paste release block.
Writes well-formed commit messages following the Conventional Commits specification. Infers the commit type and scope from the staged diff and changed file paths, asks for clarification only when the intent is genuinely ambiguous, and formats the result ready to paste into your terminal.
| name | skill-author |
| description | Guides you through designing and writing a new Claude skill from scratch — determines the right complexity tier, writes the SKILL.md with correct frontmatter and instructions, and scaffolds any supporting assets, references, or scripts the skill needs. |
| version | 1.0.0 |
| tags | ["meta","skill-authoring","developer","tooling"] |
| author | skilldex-examples |
Use this skill when the user wants to create a new Claude skill.
Before doing anything, load ../assets/skill-anatomy.md and internalize it completely. Every decision you make — frontmatter fields, section names, which directories to create — must conform to that spec. Do not invent fields or conventions that are not in it.
assets/simple-template.md — starter template for a simple skill (no supporting dirs)assets/medium-template.md — starter template for a medium skill (with assets/)assets/complex-template.md — starter template for a complex skill (with assets/, references/, scripts/)references/complexity-guide.md — decision guide for choosing the right complexity tier; load this during Step 2Understand the goal. Ask the user:
If the user has already described the skill in enough detail, skip questions you can answer yourself. Never ask for information you already have.
Choose the complexity tier. Load references/complexity-guide.md and apply it to what you learned in Step 1. State your choice and your reasoning in one sentence before continuing. If the user disagrees, discuss and adjust.
Draft the frontmatter. Fill in all required fields:
name — derive from the skill's purpose; kebab-case; must match the intended directory namedescription — one sentence starting with a verb; ≤200 chars; loaded as a tool description so make it preciseversion — always "1.0.0" for new skillstags — at least one; lowercase; hyphen-separatedauthor — ask if not providedWrite the Instructions body. Follow the section structure from skill-anatomy.md:
### Shared reference section### Bundled resources section### Workflow covering every step the skill takes### Rules section with hard invariants### Example output sectionScaffold supporting files. Based on the chosen tier:
assets/ file(s) — write out their full contents, not just a descriptionassets/ and references/ files; write a stub scripts/ shell script if a script is needed, with the correct shebang and a comment block explaining what it doesPresent the complete skill package. Output:
SKILL.md in a fenced code block labeled with the file pathThen ask: "Does this match what you had in mind, or do you want to adjust anything?"
assets/, references/, scripts/) unless the skill actually uses themskill-anatomy.mdname field in frontmatter must be identical to the directory name — flag it if the user proposes a mismatchWorkflow with steps that just say "review the output" or "make sure it's correct" — every step must be a concrete actionreferences/assets/<!-- fill in here -->