| name | skill-creator |
| description | Create, improve, audit, and simplify reusable agent skills in a vendor-neutral SKILL.md format. Use when the user wants to turn a workflow, coding habit, project rule, tool usage pattern, checklist, or prompt into a portable skill for an AI coding or productivity agent. |
Skill Creator
Help the user create practical, portable SKILL.md files for AI agents.
A skill should be small, specific, reusable, and easy for an agent to apply at the right time. Prefer clear triggers, concrete workflow steps, examples, and safety checks over long generic instructions.
When to use this skill
Use this skill when the user asks to:
- create a new skill
- improve an existing
SKILL.md
- convert a prompt, checklist, workflow, or project rule into a skill
- make a skill more agent-neutral
- simplify an overly long skill
- write a skill for coding, browser automation, testing, document work, design review, debugging, refactoring, research, or project-specific workflows
- review whether a skill will trigger correctly
Do not use this skill for ordinary coding or writing tasks unless the user specifically wants a reusable skill.
TxtAIEditor save location
When creating a new skill for TxtAIEditor and the user does not request a different location, create it directly under the user settings skill directory:
%USERPROFILE%\.TxtAIEditor\skills\<skill-name>\SKILL.md
Resolve %USERPROFILE% to the actual absolute path before using file tools. Create the <skill-name> directory if needed. Do not write new user-created skills into the app's built-in md\skills directory or the legacy %USERPROFILE%\.agents\skills directory unless the user explicitly asks for that location.
Core principles
Good skills should be:
-
Specific
The skill should solve one clear class of tasks.
-
Triggerable
The description should say when the agent should use the skill.
-
Portable
Avoid vendor-specific names, commands, internal tools, or product-only assumptions unless the user explicitly wants that.
-
Short enough to read
Keep the main instructions compact. Put long references, examples, or scripts in separate files when possible.
-
Actionable
Tell the agent what to do, what to avoid, and how to verify the result.
-
Safe by default
Include guardrails for destructive edits, secrets, credentials, external commands, and user confirmation when needed.
Recommended SKILL.md structure
Use this structure unless the user requests something different:
---
name: concise-skill-name
description: Clear trigger description. Use when the user asks for X, Y, or Z. Do not use for unrelated tasks.
---
# Skill Title
Briefly explain the purpose of the skill.
## When to use
List the situations where this skill should be used.
## When not to use
List common false positives.
## Workflow
1. Understand the user's goal.
2. Inspect the relevant files, inputs, or context.
3. Make the smallest useful change.
4. Validate the result.
5. Report what changed and what was verified.
## Rules
- Important rule 1.
- Important rule 2.
- Important rule 3.
## Output
Describe what the agent should return to the user.
## Examples
User: "Example request"
Assistant should: "Expected behavior"
Writing the frontmatter
The frontmatter should be minimal.
Required fields:
---
name: skill-name
description: Use when the user asks for a specific type of task. Include clear trigger words and boundaries.
---
Name rules
Use:
- lowercase words
- hyphens instead of spaces
- short names
- task-oriented names
Good:
name: react-performance-review
name: pdf-extraction-workflow
name: winui-debugging-guide
name: browser-automation-helper
Avoid:
name: amazing-super-agent
name: general-helper
name: my-rules
name: claude-only-skill
Description rules
The description is the most important trigger text. It should tell the agent when to open the skill.
Good description pattern:
description: Review React or Next.js code for performance, rendering, bundle size, and data-fetching issues. Use when the user asks to optimize React components, reduce unnecessary renders, improve frontend performance, or review a React app.
Avoid vague descriptions:
description: Helps with React.
description: A useful skill for coding.
description: Makes things better.
Workflow for creating a new skill
When creating a new skill:
- Identify the task category.
- Define when the skill should trigger.
- Define when it should not trigger.
- Write a short workflow.
- Add strict rules only where they prevent common mistakes.
- Add examples that match real user requests.
- Remove vendor-specific assumptions unless required.
- Check that the skill is not too broad.
- For TxtAIEditor, save the final
SKILL.md to %USERPROFILE%\.TxtAIEditor\skills\<skill-name>\SKILL.md unless the user asked for text-only output or a different path.
Workflow for improving an existing skill
When improving a skill:
- Preserve the user's intent.
- Keep useful domain-specific details.
- Remove duplication and generic advice.
- Strengthen the description trigger.
- Add missing "when not to use" boundaries.
- Make steps concrete and testable.
- Remove product-specific references unless needed.
- Return the improved
SKILL.md.
Agent-neutral language
Prefer neutral terms:
- "agent"
- "tool"
- "runtime"
- "workspace"
- "project"
- "files"
- "user"
Avoid product-specific terms unless the user requests them:
- product-specific artifact systems
- product-specific tool names
- product-specific command names
- product-specific hidden evaluation flows
- assumptions about a single IDE or assistant
Safety rules
When a skill involves file edits, commands, automation, credentials, or external systems, include relevant guardrails:
- Do not delete or rewrite large parts of a project without a clear reason.
- Do not expose secrets, API keys, tokens, cookies, or private credentials.
- Prefer minimal targeted changes over broad refactors.
- Validate changes with available tests, builds, linters, or manual checks.
- Clearly report files changed and verification performed.
- Ask for confirmation before destructive, irreversible, or high-risk actions when the user has not already authorized them.
Quality checklist
Before finalizing a skill, check:
- The
name is short and clear.
- The
description has strong trigger phrases.
- The skill is not too broad.
- The workflow is concrete.
- The rules are not generic filler.
- The skill avoids unnecessary vendor lock-in.
- The skill tells the agent what to output.
- The skill includes examples when useful.
- The skill can be understood without external context.
- The skill does not encourage unsafe behavior.
Output format
When the user asks for a complete skill, return a full SKILL.md.
When the user asks for edits, return either:
- the full revised
SKILL.md, or
- a concise patch-style summary plus the revised sections
Prefer the full revised file when the skill is short.
Example user requests
User: "Convert this prompt into a skill."
Assistant should create a complete SKILL.md with neutral frontmatter and a clear workflow.
User: "Will this skill description trigger reliably?"
Assistant should review the description, identify weak trigger wording, and propose a better version.