Sound Human — AI skill for Claude Code
Free skill for Claude/ChatGPT: build AI chat agents (DM setters, sales and support bots) that sound human in any language, by interviewing you first.
How to install Sound Human
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open setterapp/anti-ia-es and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Sound Human does
Free skill for Claude/ChatGPT: build AI chat agents (DM setters, sales and support bots) that sound human in any language, by interviewing you first. Includes prompts, examples, guardrails, tests and a checker.
Alternatives in AI
- Cl4r1t4s — LEAKED SYSTEM PROMPTS FOR CHATGPT, CLAUDE, GEMINI, GROK, PERPLEXITY, CURSOR, LOVABLE, REPLIT, AND MORE 47.2k ★
- System Prompts Leaks — Extracted system prompts from ChatGPT (GPT-5.4, GPT-5.3, Codex), Claude (Opus 4.6, Sonnet 4.6, Claude Code), G 38.6k ★
- Repomix — 📦 Repomix is a powerful tool that packs your entire repository into a single, AI-friendly file 22.7k ★
README
sound-human
A free skill for Claude, ChatGPT or Gemini that helps you build an AI chat agent (Instagram or WhatsApp setter, sales or support bot, AI SDR) whose messages read like a real person wrote them, in any language or accent.
It doesn't do magic. It interviews you, asks for your real messages and your business facts, then writes:
- the agent prompt, written as behavior with numbers instead of adjectives
- an example bank that teaches shape without getting copied word for word
- the guardrails that belong in code (leak filter, cleanup, delays, debounce)
- a test plan with synthetic customers, metrics and thresholds
It also cleans a single text (a DM, a post, an email) of AI tells.
Built from what we measured running AI setters on real Instagram DMs at [setterapp.ai](https://setterapp.ai), plus the published research and the open-source humanizer and anti-slop skills listed below.
What's inside
| File | What it covers |
|---|---|
SKILL.md |
The process: interview → voice card → decisions → write → test → hand over |
references/01-intake.md |
What to ask, in what order, and the decision list with safe defaults |
references/02-voice-from-samples.md |
Turning your real messages into a measurable voice card, in any language |
references/03-ai-tells.md |
AI tells by mechanism, with examples in English, Spanish, Portuguese, French and German |
references/04-chat-shape.md |
Messages per turn, length, questions, greetings, time, bursts, media, closing, pacing |
references/05-writing-the-prompt.md |
Prompt structure, how to write rules, layers, caching, what never goes in the prompt |
references/06-examples.md |
Why examples get copied, and how to write ones that don't |
references/07-guardrails-in-code.md |
Sanitizer, leak filter, style detector + regenerate, judge, timing, platform rules |
references/08-sales-and-support.md |
Qualifying without interrogating, price, objections, links, follow-ups, handoff |
| `references/0 |
Related Skills
LinkedIn Post Writing Skill
A complete system for writing LinkedIn posts that sound human, not AI. Built for Claude Projects. Includes ant
Professional GPT Library
90 Custom GPT system prompts for professionals — the ChatGPT equivalent of pm-claude-skills. PM, legal, financ
Offensive AI Agent Prompts
Prompts for performing tests on your Kali Linux using Gemini-cli, ChatGPT, DeepSeek, CursorAI, Claude Code, an
Open Dots
Open-source, self-hosted AI agent workspace and alternative to OpenAI Dots, Meta Muse, Grok Bot, Instinct, Man
Poka Bot
Open-source, self-hosted AI agent workspace and alternative to OpenAI Dots, Meta Muse, Grok Bot, Instinct, Man
Email Ghostwriter
An AI ghostwriter that learns your voice and writes nurture + sales emails to your list that don't sound like
Related Agents
PR Fix Agent
Fix STP/STD documents based on PR review comments from humans or bots. Classifies comments as auto-fixable or
Agent Specialist
Design, build, and optimize AI agents with strong output contracts, guardrails, and behavioral consistency. Us
Opus Answer
A simple question-answering agent that responds naturally like a standard LLM. No special research context, no