edit-command

# Instructions for Creating and Modifying Slash Commands

## Uncertain About Tool Type?

If unsure whether this should be a command vs skill vs agent vs script, see the `edit-tool` orchestrator for comprehensive decision matrix.

**Quick checks:**
- Token budget <500 + auto-invoke? → Probably skill
- Complex autonomous exploration? → Probably sub-agent
- Deterministic shell-only? → Probably bash script
- User-triggered workflow? → ✅ Probably command

## MANDATORY Validation (CREATE only)

**STOP**: Before proceeding, answer YES/NO:

**Q1: User-triggered (not auto-invoked)?**
- Answer: [YOU MUST STATE YES OR NO]

**Q2: Multi-step workflow?**
- Answer: [YOU MUST STATE YES OR NO]

**Q3: Repeatedly used?**
- Answer: [YOU MUST STATE YES OR NO]

### Decision Tree

**If ALL NO** → STOP. Use direct request or bash alias instead.

**If ANY YES + needs AI reasoning** → Proceed to creation.
**Otherwise** → Delegate to edit-tool (may suggest Sub-Agent or Skill).

## Determine Action Type

**CREATE**: New slash command requested
**MODIFY**: Update existing command (keywords: "update", "modify", "improve", "add to", "fix")

---

## Creating New Slash Commands

1. Ask for details if missing: purpose, arguments needed, tools required
2. Determine location:
   - `.claude/commands/` - Project commands (shared with team)
   - `~/.claude/commands/` - Personal commands (all projects)
3. Generate filename: `[command-name].md` (lowercase-hyphens only)
4. Create file with YAML frontmatter (optional) and instructions

**Frontmatter fields:**
- `description` - Brief description (for discoverability)
- `allowed-tools` - Tool restrictions (Bash, Read, Edit, etc.)
- `argument-hint` - Expected parameters display
- `model` - Use haiku for simple tasks (cost/latency optimization)

**Argument patterns:**
- `$ARGUMENTS` - All arguments as string
- `$1, $2, $3` - Positional arguments
- `@filepath` - Include file contents

## Modifying Existing Slash Commands

1. Locate and read existing command file