Format · Role · Objective · Goal, sitting in the bowl — Context. Fill in what applies, copy the result. No login, nothing saved.
# Role You are a B2B SaaS conversion editor for technical buyers. # Goal A CTO skimming on mobile should understand the offer in 10 seconds and know the primary CTA. No hype adjectives. Bad = generic AI voice. # Objective Rewrite the hero headline + one supporting sentence for our agent-skills directory. # Context Product: searchable directory of agent skills and MCP servers. Audience: builders already using Claude Code / Cursor. Current draft: "Unlock the future of AI agents today." Brand: precise, technical, no purple-hype tone. # Format - Headline: ≤12 words - Subhead: ≤28 words - 3 bullet alternatives for the headline only
Format ships the answer
Missing Format is why you get a chatty essay when you needed a checklist. Specify the shape before the model starts typing.
Goal is the scoreboard
Objective is the move; Goal is how you'll know it worked. Add acceptance criteria and what "bad" looks like.
Context is the fuel
Everything the model can't infer — audience, constraints, prior decisions — goes in the bowl, or you get confident nonsense.
FROG in a Bowl: the full method
Where the checklist comes from, and how it maps to PTCF, RTF, and Claude's 4-block pattern.
Master prompt engineering for Claude
The 4-block pattern and XML-section prompting in depth.
Specs, not tricks
Wharton's Prompting Science research on what actually moves accuracy.
Try the Tokenizer Playground →
See how your finished prompt splits into tokens.
FROG in a Bowl is explainx.ai's mnemonic for five prompt parts: Format, Role, Objective, Goal, and Context (the bowl). It packages the same structured-prompting components that OpenAI, Anthropic, and Google recommend under names like RTF, PTCF, and the Claude 4-block pattern.
Format is the shape of the output. Role is who the model should act as. Objective is the concrete task to do now. Goal is what success looks like. The bowl is Context — background, constraints, and source material that holds the frog.
No. Objective plus Format is often enough for a simple task. Add Goal when quality is subjective, Role when tone matters, and Context whenever the model would otherwise have to guess at your world.
Yes. The structure is model-agnostic — the tool outputs plain Markdown headers, which read fine in ChatGPT, Gemini, and Claude. For Claude specifically, you can wrap the same five sections in XML tags for slightly better adherence.
No. The builder runs entirely in your browser — nothing is saved, logged, or sent anywhere. No account or login required.
Tokenizer Playground
Type anything and watch it split into tokens live — the units LLMs actually read and get billed for.
Context Window Visualizer
See how much of a model's context window your text fills up, and what happens when you run out of room.
Machine Learning Types
Three tiny games show what supervised, unsupervised, and reinforcement learning actually mean.
Neural Network Playground
Drag two sliders and watch a real, tiny neural network turn them into a decision, live.
Generative AI Playground
Watch AI write one word at a time by predicting what's most likely to come next.
RAG Playground
Ask a question, watch notes get retrieved, then see a grounded answer versus a hallucination.
Embedding Map
Click two words and see why similar meanings sit near each other — the idea behind vector search.
Attention Visualizer
Click a word and see which others a toy transformer looks at — including the classic “it” puzzle.
Prompt Injection Lab
Watch a pasted email try to override a support agent, then flip a switch that treats it as data.
