command-creator

To create a Claude Code slash command, you author a markdown file with YAML frontmatter containing name and description, then define argument patterns and phase-structured workflow bodies that align with Claude Code loaders for argument substitution and prompt preprocessing.

Shell injection context gathering safely inlines live state like git diffs, PR data, or environment variables into a slash command prompt using correct preprocessing patterns, ensuring the model receives real-time context before executing side-effecting actions.

You add approval gates to slash commands by inserting AskUserQuestion prompts within the command body, ensuring the model pauses for user confirmation before running any destructive or side-effecting shell commands during the workflow execution phase.

You structure a multi-phase workflow by templating the command body into context gathering, planning, approval, execution, and reporting phases, so the model reliably follows each step and prompts for user input at the correct stages.

Claude Code slash command markdown files should be stored under the commands/ directory or within skill directories at skills/<name>/SKILL.md, ensuring the YAML frontmatter metadata remains discoverable by the Claude Code loaders.

Yes, you can choose free-form, positional, or named argument patterns when authoring slash commands, ensuring the argument substitutions in the markdown body match the selected pattern and align with the Claude Code preprocessing expectations.