prompt-suite

Classifies instruction-writing requests by target surface and routes them to the correct skill.

1|Updated Apr 19, 2026
One-click install
npx skills add https://github.com/fercosnt/fernando-claude-marketplace --skill prompt-suite-fercosnt
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: prompt-suite
Source: https://github.com/fercosnt/fernando-claude-marketplace/tree/main/prompt-suite/skills/prompt-suite
Command: npx skills add https://github.com/fercosnt/fernando-claude-marketplace --skill prompt-suite-fercosnt

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Requests to write instructions for Claude can target three different surfaces — CLAUDE.md files in a repository, project instructions in Claude chat or Cowork, or standalone prompts — and choosing the wrong one produces instructions that do not work where they are placed. This Skill classifies each request by surface and routes it to the right specialized skill without writing the instruction itself. ## Core Features & Use Cases - Surface classification: Maps request signals (file names, product names, keywords) to one of three surfaces: Claude Code, Project instructions, or standalone prompt. - Signal-based routing table: Directs requests mentioning CLAUDE.md or .claude/ to claude-md-generator, project or Cowork mentions to project-instructions, and prompt or N8N/Notion mentions to prompt-engineer. - Multi-skill orchestration: Handles requests spanning multiple surfaces by defining the correct execution order and passing gathered context so the destination skill does not repeat the interview. - Use Case: A user says "improve the prompt in my CLAUDE.md and set up my Cowork project" — the Skill detects both surfaces, routes to claude-md-generator first, then project-instructions, passing the confirmed context to each. ## Quick Start Ask the assistant to help you write instructions for Claude and let it determine whether they belong in a CLAUDE.md file, a project, or a standalone prompt.

Frequently Asked Questions about prompt-suite

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I decide between CLAUDE.md and project instructions?

Choose CLAUDE.md when the instruction lives in a code repository and applies to terminal or editor work. Choose project instructions when the instruction lives in a Claude chat or Cowork project and applies while that project exists.

What is the difference between a project instruction and a standalone prompt?

A project instruction persists for every conversation inside that project, while a standalone prompt applies only to a single execution such as one message, system prompt, or automation node. Permanent rules belong in project instructions; one-off tasks belong in prompts.

Can one request need both CLAUDE.md and Cowork instructions?

Yes, multi-surface requests are common, such as a code project the team also operates through Cowork. The Skill defines the order: write the CLAUDE.md first as the source, then create Cowork project instructions that point to it.

Does this Skill write the actual prompt or instructions?

No, it only classifies and routes. The actual writing is delegated to claude-md-generator, project-instructions, or prompt-engineer, and the Skill passes along the context already gathered so those skills do not restart the interview.

When should image prompts not be handled here?

Image prompts for tools like Midjourney, Flux, or Ideogram describe scenes for diffusion models rather than instructing an agent, so they fall outside this plugin. The Skill redirects those requests to deck-image-prompts instead.