Skill

prompting-craft

Generate optimized LLM prompts using the bundled prompt engineering guide and optional research. Delegates to the llm-prompt-engineer agent. Use when the user wants to create a new prompt, system instruction, or agent definition from scratch. Triggers on "craft a prompt", "write a prompt", "generate a system prompt", "create an agent prompt", "prompt for", or any request to produce a new LLM prompt.

From prove

Install

Run in your terminal

npx claudepluginhub mjmorales/claude-prove --plugin prove

Tool Access

This skill uses the workspace's default tool permissions.

Skill Content

Similar Skills

cache-components

Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.

cache-components

138.8k

claude-opus-4-5-migration

2 files

Migrates code, prompts, and API calls from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, updating model strings on Anthropic, AWS, GCP, Azure platforms.

claude-opus-4-5-migration

83.2k

bmad-brainstorming

11 files

Facilitates interactive brainstorming sessions using diverse creative techniques and ideation methods. Activates when users say 'help me brainstorm' or 'help me ideate'.

bmad-pro-skills

43.8k

Stats

Stars0

Forks0

Last CommitMar 29, 2026

Actions

View Source View Plugin View on GitHub View README

Craft Prompt

Delegate to the llm-prompt-engineer agent to generate an optimized prompt from a user's requirements.

Before Delegating

Gather these inputs from the user (ask if missing):

Target model -- which LLM will consume this prompt (Claude, GPT, open-source, general-purpose)
Task type -- classification, generation, analysis, agent behavior, system prompt, etc.
Output expectations -- what a successful response looks like
Constraints -- token budget, latency, cost sensitivity, safety requirements
Consumption context -- one-shot prompt, system instruction, or agent definition

Do NOT delegate until intent and constraints are clear.

Research Mode

Check if the user passed --research or explicitly asked for live research.

Default (no flag): Tell the agent to use the bundled references/prompt-engineering-guide.md and any cached research. Do NOT use WebSearch/WebFetch.
--research flag or user request: Tell the agent to research using WebSearch/WebFetch, then cache the results for future use.
Agent-detected gap: If the agent determines the bundled guide and cache don't cover the ask (e.g., niche model, specialized technique), it should ask the user before researching.

Delegation Instructions

Pass the gathered requirements to llm-prompt-engineer with these directives:

Read the bundled prompt engineering guide (references/prompt-engineering-guide.md)
Check the prompt research cache in priority order -- later tiers override earlier: plugin (cache/prompting/ in plugin dir), global (~/.claude/cache/prompting/), project (.prove/cache/prompting/)
Apply research mode (default: guide + cache only; --research: live web research, cache results)
Apply all relevant optimization techniques
Embed brief inline comments in the generated prompt explaining key design choices (e.g., )
Present the result with:
- Techniques applied -- each technique used and why
- Guide sections referenced -- which parts of the bundled guide informed the design
- Token estimate -- approximate token count
- Trade-offs -- what was optimized for vs deprioritized

After Generation

Ask the user if they want to:

Iterate on specific sections
Adjust the token-efficiency vs clarity balance
Add or remove techniques
Test with example inputs
Research a specific aspect they want to improve (--research)

If the user provides a file path, write the final prompt there. Otherwise, output directly.