Documentation v5.0

Welcome to Cipherc

Cipherc is a library of precision-engineered, model-specific prompt packs designed for operators who run AI at scale. Every prompt in every pack has been tested in real campaigns before shipping.

This documentation covers everything from first purchase to advanced model-specific usage. If you're new, start with Quick Start. If you want to go deep on a specific model, jump to Model Guides.

Tip: Each prompt pack is designed for a specific model. Using a Claude pack with GPT-4 will still work — but you'll get best results when you match the pack to its intended model.

Quick Start

Get up and running with your first prompt pack in under 5 minutes.

Purchase a Pack

Head to Models or Prompts, add your pack to cart, and complete checkout. You'll receive an email with a download link within seconds.

Download Your Pack

Each pack ships as a structured .md file (Markdown). Open it in any text editor, Notion, Obsidian, or paste directly into your AI interface.

Match Prompt to Model

Each prompt is labelled with its target model. Open Claude, GPT-4, Gemini, or Groq — paste the prompt as your system or user message, then fill in the [placeholders].

Iterate and Deploy

Use the output as-is or refine it. Most prompts are designed to work on the first try. For best results, keep the structure intact and only edit the bracketed variables.

Example: Using a Cold DM prompt

# COLD DM BLITZ — Prompt 03 # Target Model: GPT-4 # Category: Cold Outreach System: You are a sharp, credible crypto operator. Never sound like a bot. Never use exclamation marks. Always open with an observation, not a pitch. User: Write a cold DM to [TARGET_DESCRIPTION]. Reference: [RECENT_ACTIVITY_OR_CONTEXT]. Goal: [DESIRED_OUTCOME].

Delivery & Access

All packs are delivered digitally. Here's how it works:

Detail	Info
delivery_time	Instant — download link emailed on purchase
format	Structured Markdown (.md) — works in any editor
updates	Free forever — new prompts added to existing packs
licence	Personal use only — no resale or redistribution
refunds	Not offered on digital goods — but we'll fix any issue

Model Guides — Overview

Cipherc covers 8 models. Each has a different response profile, context window, and tone tendency. Understanding these differences is the core of getting great output.

Claude

Exceptional at long-form reasoning, nuanced tone, and multi-step tasks. Best for premium outreach, research chains, and anything requiring depth. Sensitive to instruction structure — be explicit.

GPT-4

Fast, adaptive, and great at tone matching. Excels at cold DMs, social copy, and anything requiring volume. Responds well to persona-setting system prompts.

Gemini

Strongest cross-document synthesis and narrative tracking. Use for multi-source summarisation and story development. Works best with structured input.

Groq

Fastest inference. Ideal for live Discord ops and real-time signal generation. Slightly less nuanced than Claude or GPT-4, but wins on speed when that matters.

Bolt

Code-aware and task-structured. Use for automation scripts, bot flows, and anything involving logic or workflow design.

Brave

Live-web grounded. Outputs cite or reference real-time data. Use when accuracy and source-backing matter more than stylistic polish.

Mistral

Local inference friendly. Privacy-first. No data sent to external servers. Ideal for sensitive ops, internal tooling, and air-gapped environments.

Claude — Usage Guide

Claude (Anthropic) is the recommended model for long-form, high-stakes outputs. It handles complex multi-step instructions better than any other model in the arsenal.

System Prompt Structure

You are [PERSONA]. Your tone is [TONE_DESCRIPTION]. You never [AVOID_BEHAVIOUR]. You always [TARGET_BEHAVIOUR]. Context: [RELEVANT_CONTEXT].

Key insight: Claude responds strongly to explicit behavioural constraints. Telling it what not to do (e.g. "never use buzzwords") often has more impact than telling it what to do.

Recommended Settings

Parameter	Value	Reason
temperature	0.7–0.9	Preserves creativity while staying on-task
max_tokens	600–1200	Gives Claude room for depth without bloat
system_prompt	Always set	Claude responds dramatically better with context

GPT-4 — Usage Guide

GPT-4 is the workhorse. Fast iterations, strong tone matching, and reliable output quality at scale. The Cold Outreach Engine is built entirely around GPT-4's strengths.

Persona Injection

System: You are [NAME], a [ROLE] who [CONTEXT]. You write in short sentences. You never pitch first. You lead with observations, not offers. User: [TASK]

Key insight: GPT-4 adapts its tone to match the persona you set. The more specific your system prompt persona, the more distinct and credible the output feels.

Groq — Usage Guide

Groq is the speed layer. When you need real-time outputs for live Discord drops, alert messages, or on-the-fly community responses, Groq's inference speed is unmatched.

Speed-Optimised Prompt Format

Respond in under 3 sentences. Tone: [BRIEF_TONE_DESCRIPTOR]. Task: [SINGLE_CLEAR_TASK]. Context: [ONE_LINE_CONTEXT].

Key insight: Keep Groq prompts short and task-specific. Its strength is throughput, not depth. For nuanced outputs, pair Groq's draft with a Claude refinement pass.

Prompt Library — Categories

Prompts are organised into five categories. Each category has its own optimised structure and recommended model pairings.

Cold Outreach

Cold DM Blitz (20 prompts)

Openers, hooks, closers — GPT-4 tuned

Follow-Up Machine (15 prompts)

Persistent without being pushy — Claude tuned

Influencer Pitch Kit (18 prompts)

Peer-to-peer pitch templates — GPT-4 tuned

Trading

Crypto Caller Pack (25 prompts)

Alpha calls, narratives, launch hooks — Claude

Alpha Signal Suite (30 prompts)

Real-time signal generation — Groq

Degen Playbook (22 prompts)

Full operator toolkit for volatile markets — GPT-4

Content

Twitter Thread Engine (20 prompts)

Thread structure and hooks — Claude

Long-Form Writer (15 prompts)

Newsletters, essays, research — Claude + Gemini

Automation

Discord Ops Suite (18 prompts)

Bots, welcome flows, alerts — Bolt + Groq

Community Manager Pack (16 prompts)

AMAs, onboarding, de-escalation — GPT-4

Ops

Risk & Warning Pack (12 prompts)

Rugpull warnings, scam alerts — Brave

Full Prompt Library (220+ prompts)

Every category, every pack bundled

Best Practices

A few principles that separate good prompt usage from great prompt usage:

1. Always set a system prompt

Models with a context-setting system prompt consistently outperform those without one. Even a two-sentence persona makes a significant difference.

2. Fill every placeholder

Every [BRACKETED_VARIABLE] in a prompt exists for a reason. Leaving one blank forces the model to guess — and it usually guesses wrong.

3. Match pack to model

Each pack is designed for a specific model's strengths. Cross-model usage works, but match the pack to its intended model for best results.

4. Don't over-engineer

Insight: The most common mistake is adding too many constraints. Pick 2–3 behavioural rules and let the model breathe. Over-constraining kills tone quality.

FAQ

No. You need access to the AI model the prompt is designed for — e.g. a Claude.ai account or OpenAI API key. Cipherc sells the prompts, not API access.

Yes, prompts are portable. A Claude pack will still produce good output on GPT-4 — but you'll see the biggest gains when you match the pack to its designed model.

When we update a pack, customers who purchased it receive a new download link via email automatically. No re-purchase required — ever.

Your licence covers personal use. If your whole team needs access, please reach out — we offer team licences at a fair price, not per-seat SaaS pricing.

Structured Markdown (.md). Works in Notion, Obsidian, VS Code, or any text editor. Paste directly into your AI interface or use it in your API calls.

Reach out. We'll troubleshoot with you directly. In most cases it's a placeholder left blank or a model mismatch. We stand behind the quality of every prompt.

Yes — the Mythos Master Bundle includes all 8 model packs (280+ prompts) for $29. The Full Prompt Library on the Prompts page covers all 12 prompt packs (220+ prompts) for $29.