Prompt Dictator — Patterns
Use proven input shapes to forge clean prompts you can reuse, export, and trust.
What this tool does
- Gives you repeatable input patterns for Prompt Dictator’s deterministic forge (it transforms; it doesn’t chat).
- Helps you preview/confirm safely, then interpret the A / B / C output as a usable prompt + sanity-check + optional upgrades.
- Keeps your work portable by leaning on the export bundle (Copy JSON / Download JSON) after a successful run.
What it does NOT do
- It does not guarantee what another model will output — it only forges the prompt text you feed into that model.
- It does not store your inputs privately — avoid secrets, private keys, or anything you can’t afford to expose.
- It does not replace domain expertise (legal, medical, financial, security). Use a qualified pro when it matters.
Step-by-step (the live flow)
Step 1: Paste messy input
Include the goal, audience, constraints (tone, length, format), and any must-include details. Skip passwords, private keys, or personal data you don’t want copied around.
Step 2: Preview
Use the preview to sanity-check what will be sent. If the raw input is wrong, fix it here — don’t ‘hope the tool guesses’.
Step 3: Confirm
Confirm to run the deterministic transform. If it fails, your last good export stays intact — only confirmed successes become exportable.
Step 4: Read A / B / C
A is the final prompt you’ll reuse. B is the tool’s understanding of your intent. C is optional enhancements (assumptions, follow-ups, improvements).
Step 5: Export
After a successful run, export the JSON bundle (Copy JSON / Download JSON) so you can reuse, archive, or share the exact prompt you forged.
Patterns
1) Build a spec from chaos
When to use: When you have scattered notes and need a clean spec you can hand to a builder (or future-you).
Example input
Goal: turn this into a short build spec. Notes: - users upload a PDF - extract key fields (name, email, skills) - show review screen - export JSON Constraints: no payments, must work on mobile + desktop
What you’ll get: Section A gives a ready-to-run prompt; B summarizes the intended spec; C suggests missing edge cases to add.
Pro tip: Include 3–5 ‘must not’ rules (scope locks) to keep the forged prompt from drifting.
2) Rewrite with constraints
When to use: When the content exists, but the tone/format needs hard constraints (calm, short, structured).
Example input
Rewrite this copy. Text: “We’re thrilled to announce our amazing new update!!” Constraints: - tone: calm, factual - 120 words max - include 3 bullets - no hype
What you’ll get: A produces the rewrite prompt with constraints baked in; B confirms the constraints were understood; C offers optional variants.
Pro tip: State the exact output shape you want (word limit, bullets, sections) — the tool obeys structure well.
3) Policy / guardrails prompt
When to use: When you want a reusable instruction set that keeps an AI inside boundaries (tone, refusal rules, scope).
Example input
Create a system prompt for a support assistant. Rules: - do not ask for passwords - do not claim access to user accounts - if unsure, ask one clarifying question - tone: calm, direct Output: a single system prompt + short refusal snippets
What you’ll get: A forges a clear guardrails prompt; B states the boundaries; C suggests gaps (like escalation paths and safe defaults).
Pro tip: Write refusal behavior as explicit sentences (what to say), not just ‘don’t do X’.
4) Landing page generator prompt
When to use: When you need a prompt that reliably generates landing page copy in a consistent structure.
Example input
Make a landing page prompt for: Product: weekly meal prep service Audience: busy parents Tone: practical, no hype Must include: pricing section placeholder, FAQ, and a clear CTA Output: H1 + subhead + 5 bullets + CTA + FAQ (4 Qs)
What you’ll get: A becomes your reusable generator prompt; B confirms the structure; C proposes optional sections (like objections or trust notes).
Pro tip: Add ‘avoid claims that require evidence’ to keep copy grounded and compliant.
5) Bug report → patch prompt
When to use: When you have a bug description and want a prompt that asks an AI to propose a minimal, testable patch.
Example input
Turn this into a patch prompt. Bug: - clicking Export sometimes copies empty JSON - happens after a failed run Code context: - export should only use last successful run Output: ask for root cause + minimal fix + tests
What you’ll get: A creates a patch-focused prompt; B captures the bug’s facts; C suggests test cases and failure modes to include.
Pro tip: Name the invariant explicitly (e.g., ‘export must only use last successful run’). That’s your north star.
6) Content batch prompt
When to use: When you want one prompt that produces a batch of consistent assets (titles, bullets, outlines).
Example input
Forge a prompt to generate a content batch. Topic: beginner strength training Audience: new gym members Tone: calm, practical Output: 10 titles + 10 one-sentence summaries + 5 outlines Constraints: no medical claims
What you’ll get: A becomes a batch-generation prompt; B restates constraints; C adds optional formatting checks and quality gates.
Pro tip: Ask for a fixed JSON shape in the generated content if you plan to paste it into tooling later.
7) Plain-language summary (not legal advice)
When to use: When you need a clear summary for a normal human — and you’re explicit that it’s not legal advice.
Example input
Summarize this in plain language (not legal advice). Text: - user can cancel within 14 days - refunds exclude processing fees - account deletion is permanent Output: 6 bullets + 1 short paragraph
What you’ll get: A creates a clean summarization prompt; B confirms the intended audience and constraints; C suggests clarity improvements.
Pro tip: Add ‘quote key sentences verbatim’ if accuracy matters more than paraphrase.