Author
Roo Iyer writes tested, no-preamble prompt and API recipes for PromptAttic. Every recipe ships with a receipt and a cost to test.
Stop sequences in 2026: four copy-paste recipes that cut a model off at a delimiter, bound a JSON object, force one section, and stop an agent from faking its own turn, plus where they break.
Chain of Draft keeps chain-of-thought's steps but caps each to a few words, cutting tokens hard while holding accuracy. Four recipes, the gate, and the failure mode.
Self-refine prompting in 2026: four copy-paste recipes for the draft-critique-revise loop, the paper-faithful three-call version, and the stopping rule nobody ships.
Socratic prompting in 2026: five copy-paste recipes that surface assumptions, self-interrogate in one message, debug by question, and add a maieutic consistency check, plus when to skip it.
Chain of Density is a one-call summarizing prompt that rewrites a summary five times at fixed length so it packs in the facts most 'summarize this' prompts drop. Copy-paste receipt, the failure modes, and the production variant.
Six copy-paste LLM temperature and top_p recipes for 2026: the exact setting per task type, why it works, and the failure mode it prevents.
System prompt vs user prompt in 2026: five copy-paste placement recipes with the same instruction shown in both lanes, plus the injection failure mode that makes lane choice a security decision.
ReAct prompting interleaves reasoning and tool calls: Thought, Action, Observation, repeat, until the model has enough to answer. It is how a plain prompt becomes an agent. Six copy-paste ReAct recipes below, each with the failure mode that bites first. Tested July 2026.
Zero-shot prompting is the right default for most tasks in 2026. Five recipes that make instruction-only prompts reliable, plus the exact point to add one example.
A terse, production-first prompt engineering cheat sheet for 2026: fifteen patterns, each with a one-line rule, its failure mode, and a link to the full recipe.
Five paste-ready chain-of-thought prompting examples for 2026, each with its failure mode, plus when reasoning out loud makes the answer worse.
Five paste-ready meta prompting recipes: make the model write, critique, rubric, and template your prompts, each with its failure mode. Plus the honest part: when meta prompting just wastes tokens.
