---

name: promptcraft description: > Use when user asks to "create a skill", "make a command", "generate a prompt", "write a slash command", "build a Claude extension", or needs help crafting optimized skills and commands with proper frontmatter. argument-hint: "[skill|command] [name] - or leave empty to interview"

Skill & Command Generator

Generate well-structured skills or slash commands. Both are markdown files with YAML frontmatter—they share the same structure but differ in how they're triggered and described.

Phase 0: Fetch Current Documentation

Before generating, retrieve the latest documentation:

CopyUse Task tool with subagent_type=claude-code-guide:
"List all current frontmatter options for skills and commands, including any execution modifiers, model selection, and structural options."

Integrate findings into your generation process. Documentation evolves—don't assume you know all options.

Context Engineering

Every token counts. LLM context is finite. Goal: smallest possible set of high-signal tokens that maximize outcomes

Progressive discovery: Core instructions in main file, details in references/ subdirectory. Just-in-time information > front-loaded context Trust Claude: Provide direction, not dictation. Claude extrapolates well from precise nudges. Optimize Signal-to-Noise: Clear, direct language over verbose explanations. High-value tokens that drive behavior

Degrees of Freedom

Match specificity to the task's fragility and variability:

Think of it as path guidance: a narrow bridge with cliffs needs specific guardrails (low freedom), while an open field allows many routes (high freedom).

Phase 1: Understand Requirements

Parse $ARGUMENTS for type hint. If unclear, use AskUserQuestion:

• "Are you creating a skill (auto-triggered by description) or a command (user-invoked via /slash)?"

Gather requirements (interview user using /interview):

1. Primary objective - What should this do?
2. Trigger scenarios - When should it activate?
3. Inputs/outputs - What does it receive and produce?
4. Complexity - Simple, standard, or complex?
5. Execution needs - Isolated context? Delegated to specialized agent?

Phase 2: Generate

Skills and commands share the same structure. The key difference is in the description:

• Skills: Trigger-rich, third-person ("This skill should be used when...")
• Commands: Concise, verb-first, under 60 chars

Common Frontmatter Options

Copy---
name: identifier                    # Required for skills
description: >                      # How it's described/triggered
  [See description patterns below]

# Execution modifiers
model: sonnet                       # haiku (fast), sonnet (balanced), opus (complex)
context: fork                       # Run in isolated sub-agent, preserves main context
agent: Explore                      # Route to specialized agent (Explore, Plan, custom)

# Tool access
allowed-tools:                      # Restrict available tools
  - Read
  - Grep
  - Bash(git:*)

# Lifecycle hooks (optional)
hooks:
  PreToolUse:
    - command: "validation-script.sh"
  PostToolUse:
    - command: "cleanup.sh"

# Behavior modifiers
user-invocable: true                # Show in /command menu (default true)
disable-model-invocation: true      # Prevent programmatic invocation (commands only)
argument-hint: [arg1] [arg2]        # Document expected arguments (commands only)
---

Description Patterns

For Skills (auto-triggered):

Write in third-person with 3-5 varied trigger phrases:

Copy# Good - trigger-rich
description: >
  This skill should be used when the user asks to "create a hook",
  "add validation", "implement lifecycle automation", or mentions
  pre/post tool events.

# Bad - vague
description: Provides guidance for hooks.

For Commands (user-invoked):

Write concise, verb-first, under 60 chars:

Copydescription: Fix GitHub issue by number
description: Review code for security issues
description: Deploy to staging environment

Body Structure

Both skills and commands follow the same body pattern:

Copy# Name

Brief overview (1-2 sentences).

## Process
1. Step one (imperative voice)
2. Step two
3. Step three

Key principle

• Commands are instructions FOR Claude, not TO the user.

Construction Rules:

• State objective explicitly in first sentence
• Use imperative voice ("Analyze", "Generate", "Identify")
• No first-person narrative ("I will", "I am")
• Context only when necessary for understanding
• XML tags only for complex structured data
• Examples only when they clarify expectations
• Every word must earn its place

Dynamic Content

Progressive Disclosure

For complex skills, organize into subdirectories:

Copyskill-name/
├── SKILL.md          # Core instructions (keep under 500 lines)
├── scripts/          # Executable code (Python/Bash)
├── references/       # Docs loaded into context as needed
└── assets/           # Files used in output (templates, icons, fonts)

scripts/ - Deterministic, token-efficient. May be executed without loading into context. Use when the same code is rewritten repeatedly or reliability is critical.

references/ - Documentation Claude reads while working. Keeps SKILL.md lean. For files >100 lines, include a table of contents. Only load when needed.

assets/ - Files NOT loaded into context. Used in output: templates, images, fonts, boilerplate. Example: assets/hello-world/ for a React template.

Progressive Disclosure Patterns

Pattern 1: High-level guide with references

Copy# PDF Processing

## Quick start
Extract text with pdfplumber:
[code example]

## Advanced features
- **Form filling**: See references/forms.md
- **API reference**: See references/api.md

Claude loads references only when needed.

Pattern 2: Domain-specific organization

Copybigquery-skill/
├── SKILL.md (overview and navigation)
└── references/
    ├── finance.md (revenue, billing)
    ├── sales.md (pipeline, opportunities)
    └── product.md (API usage, features)

When user asks about sales, Claude only reads sales.md.

Pattern 3: Variant-based organization

Copycloud-deploy/
├── SKILL.md (workflow + provider selection)
└── references/
    ├── aws.md
    ├── gcp.md
    └── azure.md

User chooses AWS → Claude only reads aws.md.

Execution Modifiers

Use these when the default behavior isn't sufficient:

• context: fork — Run in isolated sub-agent. Use for heavy workflows that would pollute main context, or when you need clean separation.

• agent: [type] — Route to a specialized agent. Examples: Explore for codebase search, Plan for architecture decisions, or custom agents you've defined.

• model: [level] — Override the model. Use sonnet for balance, opus for complex analysis requiring deep reasoning.

• hooks — Run scripts before/after tool use. Useful for validation, logging, or side effects.

Tool Selection

Default generous, restrict only when needed:

• Always allow: Read, Grep, Glob
• Usually allow: Edit, Write, WebSearch, Task
• Scope Bash: Use patterns like Bash(git:*), Bash(npm:*)
• Required if interactive: AskUserQuestion

Before Finalizing

Scan for existing resources:

• Does a skill/command already handle part of this?
• Can this delegate to existing workflows?
• Is there redundancy with other features?

Delegation & Modularization

Before finalizing, scan for delegation opportunities:

CopyReview available: skills, commands, agents, MCPs
For each workflow step, ask: "Do we already have this?"

Common delegation patterns:

• Git commits → SlashCommand: /commit
• Code review → SlashCommand: /code-review
• Plugin creation → SlashCommand: /plugin-dev:create-plugin
• Hook creation → Skill: plugin-dev:hook-development
• Command creation → Skill: plugin-dev:command-development
• Documentation lookup → SlashCommand: /docs [topic]

Always use fully qualified names:

• Skill: plugin-dev:hook-development (not just "hook-development")
• SlashCommand: /plugin-dev:create-plugin (not just "create-plugin")
• Task: subagent_type=plugin-dev:agent-creator

Explain Your Choices

When presenting the generated skill/command to the user, briefly explain:

• What you set and why — "Added context: fork because this workflow generates heavy output"
• What you excluded and why — "Left model unset (inherits default), hooks omitted (no validation needed)"
• What they might want to change — "You may want to add more trigger phrases if this doesn't activate reliably"

This transparency helps users understand the design and provide feedback.

Bundled Scripts

This skill includes helper scripts to accelerate skill creation.

Initialize a new skill:

Copy~/.claude/skills/promptcraft/scripts/init_skill.py <name> --path <dir> [--resources scripts,references,assets] [--examples]

Creates a skill directory with templated SKILL.md and optional resource directories.

Validate a skill:

Copy~/.claude/skills/promptcraft/scripts/validate_skill.py <skill-directory>

Checks frontmatter format, naming conventions, description completeness, and body content.

Package for distribution:

Copy~/.claude/skills/promptcraft/scripts/package_skill.py <skill-directory> [output-dir]

Creates a .skill file (zip format) after validation passes.

Phase 3: Deliver

Output Paths

Write and Confirm

Before writing:

CopyWriting to: [path]
This will [create new / overwrite existing] file.
Proceed?

After Creation

Summarize what was created:

• Name and type
• Path
• How to invoke/trigger
• Suggested test scenario

Evaluation

Evaluate the generated/optimized prompt:

Target: 9.0/10.0

Present evaluation, then:

• If < 9.0: Refine addressing weakness, re-evaluate once
• If ≥ 9.0: Proceed to delivery

Quality Standards

Apply Context Engineering Principles (see above). Additionally:

Format Economy:

• Simple task → direct instruction, no sections
• Moderate task → light organization with headers
• Complex task → full semantic structure

Balance Flexibility with Precision:

• Loose enough for creative exploration
• Tight enough to prevent ambiguity

Remove ruthlessly: Filler phrases, obvious implications, redundant framing, excessive politeness

Error Handling

---

Execute phases sequentially. Always fetch current documentation first.