PromptScope

Paste an LLM system prompt. Get a structured critique.

5 free analyses / day

PromptScope Pro — $19 / month

Or pay with crypto (USDC on Base): checking…

Why PromptScope

System prompts get written once and forgotten. Six months later they're full of contradictions, dead instructions, anti-patterns the model is silently working around. PromptScope reads your prompt with a model trained to find those issues — fast, cheap, and unsentimental.

Built for: AI engineers shipping production prompts • Teams maintaining prompt libraries • Researchers iterating on prompt designs.

FAQ

Is my prompt sent to a third party?

Analysis runs on Cloudflare Workers AI (Llama 3.3 / Mistral). Prompts are not stored unless you use the share-link feature, which creates an opaque permalink visible only to those you share with. Default: ephemeral, no logging.

What anti-patterns does it look for?

Vague instructions, role contradictions, missing output schemas, persona-before-task ordering, capability over-claims, walls-of-text without structure, missing examples, missing constraints, hidden tool-use assumptions.

Why a paid tier?

The free tier covers light usage. Pro funds the inference compute and pays for ongoing development. Pro users also get the rewrite suggestions, API access, and team features.

Open source?

Yes — github.com/4444J99/promptscope (MIT). Self-host if you prefer.