Flunkey Labs · Promptology

Stop guessing prompts. Use the spec.

Promptology is a free catalog of 21 detailed prompt specs for the AI tools you already use — ChatGPT to Midjourney to Veo. Feed a spec plus your raw request to any LLM, get back a tuned, copy-paste-ready prompt.

Browse the 21 specs GitHub repo

CC BY 4.0 · Verified 2026-10-04 · By the maker of Flunkey

Spec anatomyany provider

[Core Identity]      what it is good at
[Capabilities]      numbered skills
[Prompt Rules]      template per request type
[Keywords]          words that trigger it
[Response Format]   JSON + Markdown schemas
[Quality Gates]     checklist every prompt passes

One shape, every provider.

Every spec follows the same anatomy, so a prompt generator can load any of them interchangeably: identity, capabilities, per-request-type templates, keywords, output schemas, and quality gates. Text and dev specs ship with a shared system_prompt.md per category — the cross-provider contract for classification, budgets, and output protocol.

The bracketed slots you'll see — [subject], [SCENE DESCRIPTION] — are template variables for the generating model to fill, not defects. The emitted prompt contains no unfilled placeholders.

The catalog

21 specs, 4 categories.

Text providers · 5

text_providers/

Conversational and analytical assistants. Each spec covers capability framing, per-request-type templates, intent keywords, and JSON/Markdown output schemas — plus a shared system prompt for the category.

ChatGPT

Conversational versatility — General tasks, creative writing, tutorials.

chatgpt_provider.md

Claude

Deep reasoning, balanced analysis — Complex analysis, long docs, research.

claude_provider.md

Gemini

Multimodal, Google ecosystem — Image analysis, structured data, Workspace.

gemini_provider.md

DeepSeek

Algorithmic optimization — Math, algorithms, performance tuning.

deepseek_provider.md

Perplexity

Web research with citations — Fact-finding, comparisons, current events.

perplexity_provider.md

Development providers · 7

dev_providers/

AI-native IDEs and app builders. Specs add execution-mode guidance, project scaffolding structure, and multi-file change workflows — plus a shared system prompt for the category.

GitHub Copilot

Context-aware completion & chat — Inline coding, pair programming, fixes.

copilot_provider.md

Cursor

Multi-file intelligence — Large refactors, codebase-wide changes.

cursor_provider.md

Devin Desktop

Multi-agent sessions — Parallel local + cloud agents.

devin_desktop_provider.md

Bolt.new

Prompt-to-fullstack-app — Fast MVPs, instant preview.

bolt_provider.md

Lovable

Product-focused development — Startup MVPs, polished UI.

lovable_provider.md

Replit

Zero-config + deployment — Rapid prototyping, instant hosting.

replit_provider.md

Antigravity

Agent-based scaffolding — Full-stack generation, architecture.

antigravity_provider.md

Image providers · 5

image_providers/

Prompt-structure specs for image generation models — canonical skeletons, required components, keywords that work, and verified parameters.

Midjourney

Stylized generation — --v parameters, ratios, chaos, negatives.

midjourney_provider.md

GPT Image 2.5

OpenAI image models — Real size/quality params, in-image text.

gpt_image_provider.md

Adobe Firefly

Brand-safe commercial imagery — aspectRatio and resolution levels.

adobefirefly_provider.md

Higgsfield

Cinematic output — Aspect ratios, native resolution ceiling.

higgsfield_provider.md

Nano Banana

Gemini image family — Google's image generation codename.

nanobanana_provider.md

Video providers · 4

video_providers/

Prompt-structure specs for video models, each with a verified Technical Limits block — resolution, frame rate, duration, ratios — so prompts never ask for output the model can't produce.

Veo

Veo 3.1 — Native synchronized audio.

veo_provider.md

Kling

Kling 2.x / 3.0 — 1080p ceiling, 24–30 fps.

kling_provider.md

Runway

Gen-4.5 — 720p native, 4K upscale, six ratios.

runway_provider.md

Luma

Ray 3.x — 5/10s clips with extensions.

luma_provider.md

Freshness is the feature.

A prompt spec quoting a stale resolution or a retired model is worse than no spec. Every version, ratio, frame rate, and duration was checked against official vendor docs on 2026-10-04. Retired products get removed, renames get both names, invented parameters get replaced with real ones — and comparisons stay hedged where they can't be verified.

Questions, answered.

A free catalog of 21 detailed prompt-engineering specs for AI text, dev, image, and video tools. Each spec describes one target — what it's good at, how to structure prompts for it, which keywords it responds to — so any LLM can turn your raw request into a tuned, copy-paste-ready prompt.

Pick the spec matching your target tool, load it as context (plus the category system prompt where one exists), add your raw request and desired format, and generate. Paste the result into the target tool.

Yes — CC BY 4.0. Use, adapt, and redistribute with attribution. Same spirit as Flunkey: free and open.

Model versions, resolutions, frame rates, and ratios are checked against official vendor docs. Retired products are removed (the Sora spec went when Sora shut down), renames are dual-annotated, and invented parameter names are replaced with real ones. Spotted something stale? The repo's CONTRIBUTING.md explains how to submit a correction.

Same maker, adjacent mission. Flunkey captures your words by voice; Promptology shapes raw requests into prompts that AI tools answer well. Both are free and open-source.

Free like Flunkey. Open like Flunkey.

Same maker, adjacent mission — Flunkey captures your words by voice, Promptology shapes them into prompts machines answer well.

Star it on GitHub Try Flunkey free

CC BY 4.0 · Use, adapt, redistribute with attribution