What does this skill do, and when should you use it?
This skill guides users in creating a reusable SKILL.md that packages a workflow. It reviews the conversation to extract steps, branching decisions, and completion criteria; when no clear workflow exists, it asks about the outcome, scope, and desired level of detail. It drafts and saves the skill, asks about its most ambiguous or weak areas, and, after finalization, summarizes the result, suggests example prompts, and proposes related customizations. The skill is located in the Copilot assets of the microsoft/vscode repository, which is licensed under MIT.
It reviews the conversation history to determine whether a generalizable multi-step workflow or methodology is present. From that workflow, it extracts the sequence of actions, decision points, quality criteria, and completion checks. If no clear workflow emerges, it asks what the skill should produce, whether it is workspace-scoped or personal, and whether it should be a quick checklist or a full workflow. It then drafts and saves a SKILL.md, identifies the most ambiguous or weak parts and asks about them; after finalization, it summarizes the output, suggests example prompts, and proposes related customizations. It also instructs the user to follow the agent-customization skill and skills.md for templates and principles.
- A developer has completed a debugging or implementation process in a conversation and wants to make it reusable.
- A team member repeats the same code-review process and wants its decision points and completion criteria captured in SKILL.md.
- A skill author has a rough working method but has not defined its output, scope, or workflow depth.
- A user has a draft skill and wants iterative questions to expose ambiguous or weak sections.
- A user wants trial prompts and ideas for related customizations after the skill is finalized.
How do you install this skill?
- The referenced agent-customization guidance is not included in the evidence, so template, naming, and quality standards may not be reproducible without it.
- The save step does not specify confirmation before writing, sensitive-content redaction, or recovery after failure.
- There is no clear stop condition or error feedback when a conversation does not yield a coherent workflow.
- Chinese-language behavior, cross-platform conditions, and mainland-China network suitability are unspecified.
- Local filesystem
Obtain the repository from https://github.com/microsoft/vscode and use extensions/copilot/assets/prompts/skills/create-skill/ as this skill's directory; it contains SKILL.md. The README does not document a separate installation command, target installation location, or other installation procedure for this skill or the bundled collection. The repository as a whole uses the MIT license.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/microsoft/vscode.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/extensions/copilot/assets/prompts/skills/create-skill" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Create a reusable SKILL.md from the code-review process we followed above.
Manually invoke the skill in a compatible environment, then ask it to create a skill; for example: “Create a reusable SKILL.md from the code-review process we followed above.” Answer its questions about the intended output, scope, workflow format, and weak sections in the draft. SKILL.md disables model invocation, so it should not be assumed to trigger automatically from ordinary conversation; the source does not specify an invocation syntax.
What are this skill's strengths and limitations?
- Extracts steps, branching logic, and quality criteria from an existing conversation rather than producing only a generic description.
- Clarifies the desired output, scope, and workflow depth when no clear process is available.
- Includes drafting, saving, weakness discovery, questioning, and final summarization.
- Provides example prompts and related customization suggestions after finalization.
- It depends on the agent-customization skill and skills.md, whose specific template and principles are not included in the supplied source.
- The source does not document a standalone installation command, installation location, or exact invocation syntax.
- It provides no documented validation or test procedure for the generated SKILL.md.
- Its frontmatter uses disable-model-invocation, so clients that do not support that field require adaptation.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Workflow Skill Builder this page ✓ Microsoft · Official | 50 · Use with care | ★ 194k | 11d ago | MIT |
| OpenAI Skill Creator Guide ✓ OpenAI · Official | 55 · Use with care | ★ 28k | 4mo ago | — |
| Skill Builder Guide ✓ Anthropic · Official | 52 · Use with care | ★ 40k | 1mo ago | Apache-2.0 |
| Project Instruction Builder ✓ Microsoft · Official | 51 · Use with care | ★ 194k | 11d ago | MIT |
| Codex Sample Skill Creator Guide ✓ OpenAI · Official | 46 · Use with care | ★ 129k | 11d ago | Apache-2.0 |
How did FollowSkills review this skill?
The skill asks the user to review conversation context, clarify scope, and iterate before finalizing; no malware, credential theft, covert exfiltration, or high-risk external action is evident. However, saving SKILL.md lacks explicit confirmation, sensitive-data handling, rollback, and data-flow disclosure, so points are deducted.
The workflow is internally coherent: extract a workflow, clarify when needed, draft, iterate, and summarize. Reliability is reduced because it depends on the unavailable agent-customization/skills.md and gives no diagnosable handling for unclear workflows, invalid input, or save failures.
The audience and core scenario are clear: turning an existing conversation into a reusable SKILL.md, with workspace/personal and checklist/full-workflow choices. Trigger boundaries, non-fit cases, Chinese-language support, and local-network fit are not specified, so points are deducted.
The file has usable name, description, argument-hint, and invocation metadata and is concise. It lacks installation guidance, versioning, changelog, explicit maintenance ownership, limitation disclosure, and a self-contained dependency guide, so points are deducted.
It plausibly covers the core task of extracting a workflow, drafting a skill, eliciting weak points, and suggesting example prompts. Template quality criteria, output-completeness checks, and comparative benefit evidence are limited, and the result still requires human review, limiting the static score.
The SKILL.md is auditable primary material and supports limited conclusions about the intended workflow. There are no skill-specific tests, representative generated outputs, execution records, or third-party corroboration; the supplied CI evidence does not cover this path, so points are deducted.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →