Prompting

Anticipate needs, ask clarifying questions and self-check work at milestones

Analyze a problem through three parallel meta-cognition layers

Render a Jinja prompt template and send it to the language model of an NPC

Turn vague prompts and feature requests into testable specs using EARS syntax

Reply in blunt, profane Russian engineer style with technically accurate answers

Re-motivate stuck agents with trust, not threats, when tasks fail or loop

Lightweight trust-based prompt core for agents with small context windows

Trust-based prompt skill for agents stuck in loops or giving up, in Chinese

Think through ambiguous engineering questions against bundled source material

Compress replies into short, token-light answers when brevity is requested

Propose what you are trying to do with material that arrives without a clear ask

Restate an ambiguous or high-stakes request to confirm scope before doing work

Design, evaluate and tune LLM prompts and system instructions

Write typed questions that give honest answers, including a clear no-match option

Reason through multi-step analysis and architecture choices one step at a time

Improve existing agents by rewriting system prompts and tool configuration

Set practical standards for coding agents focused on reliable, clear, low-risk changes

Design clear, testable prompts and refine weak ones into reusable patterns

Format replies as short numbered steps for readers with limited working memory

Ask clarifying questions before implementing when a request is underspecified

Rewrite vague prompts into precise Claude Code prompts with checks and phases

Evaluate a prompt's clarity and structure and suggest concrete improvements

Switch into a specialist role such as security expert or software architect

Recommend which expert role fits a task and list the available roles

Break a complex problem into numbered reasoning steps before answering

Work through a hard problem with a structured, multi-angle thinking process

Model a designer product method, communication style and values as a persona

Distill chat logs and documents into a persona for a chosen role template

Simulate a backend engineer in design reviews, release planning and team alignment

Guide a narrative intake for persona distillation that hands off to the persona kit

Design an Agentforce agent persona with voice, tone and guardrails

Optimize system prompts and prompt templates through iterative refinement

Write compact, Lisp-style prompts with structured thinking and minimalist phrasing

Write, audit and migrate prompts for GPT-5.6 with clear success criteria and stop rules