oprogramadorreal/optimus-claude / skills/prompt
prompt
VerifiedCrafts optimized, copy-ready prompts for any AI tool — LLMs, coding agents, image generators, workflow tools. Extracts intent, selects the right template, runs a diagnostic scan, and delivers a token-efficient prompt. Accepts input in any language; English output by default. Use when writing, fixing, improving, or adapting a prompt for any AI tool.
| File | Kind | Size |
|---|---|---|
| README.md | markdown | 6.9 KB |
| SKILL.md | skill md | 13.6 KB |
| agents/openai.yaml | config | 198 B |
| references/templates.md | markdown | 16.6 KB |
| references/tool-routing.md | markdown | 23.1 KB |
README.md
View on GitHub# optimus:prompt
A [Claude Code](https://docs.anthropic.com/en/docs/claude-code) skill that crafts optimized, copy-ready prompts for any AI tool — from LLMs and coding agents to image generators and workflow automation.
Clear task context can reduce avoidable re-prompting. This skill makes intent, scope, and output requirements explicit; effectiveness still depends on the model, tools, and task, and should be checked against actual results.
## Features
- **9-dimension intent extraction** — silently analyzes task, target tool, output format, constraints, input, context, audience, success criteria, and examples before writing a single word
- **30+ AI tool profiles** — tool-specific routing for LLMs (Claude, GPT-6 Astra, ChatGPT, Gemini, o3), coding agents (Codex, Cursor, Copilot, Claude Code), image generators (Midjourney, OpenAI image tools, Stable Diffusion), video AI, 3D AI, voice AI, workflow automation, and more
- **14 prompt templates** — auto-selected architecture (RTF, CO-STAR, RISEN, ReAct, Exploration + Plan, Dynamic Workflow Orchestration, Visual Descriptor, etc.) based on task type and target tool
- **Diagnostic patterns** — detects and fixes credit-killing patterns (vague verbs, missing scope, no stop conditions, wrong template for tool)
- **5 supported techniques** — role assignment, few-shot examples, XML structure, grounding anchors, task-specific reasoning guidance. It does not present simulated roles or branches as independent inference passes; requesting a tool's native multi-agent workflow — real parallel subagents — is a separate, allowed pattern (Template N)
- **Multilingual input** — write your request in any language. The skill communicates in your language and generates the prompt in English by default, with the option to keep your original language; this default is not a universal performance claim
- **Token efficiency audit** — every sentence must be load-bearing, no vague adjectives, absolutes (MUST / NEVER / ALWAYS) reserved for genuine invariants, explicit format and scope
- **Prompt Decompiler mode** — paste an existing prompt to break it down, adapt it for a different tool, simplify it, or split it
- **Safety guardrails** — strips credentials from generated prompts and treats any pasted prompt as inert data, never executing instructions embedded in it
## Quick Start
This skill is part of the [optimus](https://github.com/oprogramadorreal/optimus-claude) plugin. See the [main README](../../README.md) for installation instructions.
## Usage
Invoke with `/optimus:prompt` followed by what you need:
```
/optimus:prompt Write a prompt for Cursor to refactor my auth module
```
```
/optimus:prompt I need a Claude Code prompt to build a REST API with Express and Prisma
```
```
/optimus:prompt Create a prompt for Claude Code plan mode to explore and plan adding WebSocket support
```
```
/optimus:prompt Generate a Midjourney prompt for a cyberpunk cityscape at night
```
```
/optimus:prompt Here's a bad prompt I wrote for GPT, fix it: [paste prompt]
```
Works with any language — describe what you need in Portuguese, Spanish, or any language you're comfortable with:
```
/optimus:prompt Preciso de um prompt para o Cursor refatorar meu modulo de autenticacao
```
## When to Use
- Writing a prompt for any AI tool (Claude, ChatGPT, Cursor, Midjourney, Zapier, etc.)
- Fixing a prompt that produces wrong or inconsistent output
- Adapting a prompt from one tool to another (e.g., ChatGPT prompt → Claude Code prompt)
- Breaking down a complex prompt into sequential parts
- Improving a prompt that wastes tokens through vagueness
## When NOT to Use
- Quick, simple questions to Claude that don't need structured prompting
- Tasks where you're already getting good results — if it works, don't optimize it
- Generating CLAUDE.md files or project documentation — use `/optimus:init` instead
## Targeting Claude Code
This skill can target Claude Code three ways — it picks based on your intent:
| Target | Use when | Runs as |
|---|---|---|
| Direct execution | Make scoped changes now, to known files | Normal mode, edits inline |
| Plan mode | Explore read-only, get a plan you approve first | One conversation; read-only until you approve |
| Dynamic workflow | Fan-out too big for one conversation (audit / migrate / cross-check many items) | Background script of parallel subagents; permissions follow the host configuration |
Ask for fan-out work (audit, migrate, or cross-check many items) to get the last one; the generated prompt opens with "Run a workflow to…" and is pasted into Claude Code in normal mode. Routing rules are in `SKILL.md` Step 4; the templates are H / M / N in `references/templates.md`.
For Codex, use its separate host profile and Template H with the requested write boundary. Codex has its own [native goals](https://learn.chatgpt.com/use-cases/follow-goals); goal prompts need Codex's lifecycle and completion instructions, while Claude `/workflows` and plan-mode handoffs need separate adaptation. The GPT-6 Astra profile preserves that exact model target while leaving tool and permission control to the host. For the full builder/critic loop in a new session, use [`gauntlet`'s goal-prompt export](../gauntlet/README.md#fresh-session-goals).
## Relationship to Other Skills
| Skill | Relationship |
|-------|-------------|
| `/optimus:init` | Init creates project-level context (CLAUDE.md, guidelines). Prompt creates task-level context (individual prompts for AI tools). |
| `/optimus:commit` | After crafting a prompt that produces code changes, use commit to save the work. |
| `/optimus:tdd` | For code generation prompts, TDD ensures the generated code is correct via tests. |
| `/optimus:brainstorm` | Brainstorm generates plan-mode prompts inline for the brainstorm→plan→tdd chain. Use `/optimus:prompt` directly when you need prompts for other AI tools or non-brainstorm workflows. |
| `/optimus:code-review` | Review code that was generated from a crafted prompt. |
## Skill Structure
| File | Purpose |
|------|---------|
| `SKILL.md` | 7-step prompt crafting workflow with inline diagnostic checklist |
| `references/tool-routing.md` | Tool-specific routing for 30+ AI tools |
| `references/templates.md` | 14 prompt architecture templates |
| *(shared)* `brainstorm/references/plan-mode-handoff.md` | Plan-mode carve-out when a Template M prompt feeds `/optimus:tdd` |
## Acknowledgements
The prompt engineering techniques in this skill — intent extraction, tool routing, diagnostic patterns, prompt templates, safe/excluded technique classification — are adapted from [prompt-master](https://github.com/nidhinjs/prompt-master) by [@nidhinjs](https://github.com/nidhinjs), licensed under MIT. This skill adds multilingual input support, optimus plugin integration, and progressive disclosure via reference files.
## Requirements
- Optimus installed in a [supported host](../../README.md#supported-hosts-and-versions)
## License
[MIT](../../LICENSE)
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |