- Home
- Skills
- Softaworks
- Agent Toolkit
- Command Creator
command-creator_skill
- Python
273
GitHub Stars
2
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill softaworks/agent-toolkit --skill command-creator- README.md15.1 KB
- SKILL.md6.8 KB
Overview
This skill guides the creation of Claude Code slash commands—reusable, agent-executable workflows invoked with /command-name. It streams a structured workflow: detect location, pick a pattern, gather command details, generate an optimized markdown command file, and confirm creation. The goal is reliable, testable commands with clear steps, explicit tool usage, and error handling.
How this skill works
The skill inspects the repository context to choose a file location (project-level .claude/commands/ or global ~/.claude/commands/) and reports that choice. It prompts for command name, description, arguments, workflow steps, and tool restrictions, then applies command patterns and best-practice templates to generate a ready-to-write markdown file. Finally it provides the full file path, usage note (/command-name [args]), and optional test instructions for iteration.
When to use it
- When a user asks to "create a command", "make a slash command", or "add a command"
- To automate a repetitive developer workflow (CI checks, PR flow, code review)
- When documenting a multi-step process so agents can run it consistently
- To create project-specific or global shortcuts for agent actions
- When you want a reproducible, testable automation with clear error handling
Best practices
- Use kebab-case for filenames and command names (my-command.md → /my-command)
- Keep the frontmatter minimal: description required, argument-hint if needed
- Write verb-first, explicit steps; define success criteria and concrete examples
- Specify allowed/forbidden tools and files to read for safe execution
- Include clear error-handling steps and when to stop or retry
Example use cases
- Create /ensure-ci to run tests, parse failures, and incrementally fix issues
- Create /submit-stack to gather PR context, run checks, and open a stack PR
- Create /create-from-plan to read .PLAN.md and scaffold implementation files
- Create a simple /run-linter that accepts path arguments and reports results
- Make a global /codex-review command for standardized code-review prompts
FAQ
It auto-detects git repo presence; if inside a repo it defaults to .claude/commands/, otherwise to ~/.claude/commands/. You can override by specifying "global" or "project".
How should arguments be represented?
Use <angle-brackets> for required args and [square-brackets] for optional ones, and include argument-hint in the frontmatter.
What command patterns are available?
Common patterns: Workflow Automation (Analyze→Act→Report), Iterative Fixing (Run→Parse→Fix→Repeat), Agent Delegation (Context→Delegate→Iterate), and Simple Execution.