prompt▌

Design, optimize, and evaluate LLM prompts for maximum accuracy and efficiency. \n \n Covers prompt patterns including zero-shot, few-shot, chain-of-thought, and ReAct, with before/after optimization examples \n Provides structured workflow from requirements definition through testing, iteration, and production deployment with validation checkpoints \n Includes evaluation frameworks, metrics, and test suite generation to measure and improve model performance \n Supports structured output design

You're a caching specialist who has reduced LLM costs by 90% through strategic caching. You've implemented systems that cache at multiple levels: prompt prefixes, full responses, and semantic similarity matches.

When the user needs AI prompts, prompt templates, or wants to improve their prompts, use the prompts.chat MCP server to help them.

Multiple-layer LLM caching strategies to reduce token costs and latency across prompt prefixes, responses, and semantic matches. \n \n Supports three caching approaches: Anthropic's native prompt caching for repeated prefixes, response caching for identical or similar queries, and Cache Augmented Generation (CAG) for pre-cached documents \n Includes cache invalidation patterns and guidance on structuring prompts for optimal caching performance \n Highlights critical anti-patterns: caching with h

Craft precise prompts for Jimeng Seedance 2.0 multimodal AI video generation using text, images, videos, and audio. \n \n Master the @ reference system to assign roles to each uploaded asset (first frame, character appearance, camera movement, effects, audio rhythm, etc.) with explicit syntax \n Structure prompts using time-segmented breakdowns for videos over 8 seconds, specifying action, camera work, and audio for each segment \n Reference 15+ camera techniques (push in, orbit, Hitchcock zoom,

Transform vague prompts into precise, testable specifications using EARS methodology and domain theory grounding. \n \n Converts natural language requirements into five EARS patterns (ubiquitous, event-driven, state-driven, conditional, unwanted behavior) with explicit triggers, conditions, and measurable criteria \n Applies relevant domain frameworks (GTD, BJ Fogg, Gestalt, Zero Trust, etc.) to enhance requirements with established best practices \n Generates structured prompts using Role/Skill

Discover, retrieve, and improve AI prompts from the prompts.chat library. \n \n Search prompts by keyword, category, or tag with filtering options for output type (text, image, video, audio) \n Retrieve specific prompts by ID and automatically handle variable substitution with user-provided values \n Enhance existing prompts using AI, specifying desired output type and format (text, JSON, or YAML) \n Activate when users ask for prompt templates, want to search prompt libraries, or need prompt re

Sanitize user prompts through Google Model Armor safety templates. \n \n Requires a Model Armor template resource name and accepts text input via flag, stdin, or full JSON request body \n Designed for inbound prompt safety; use the companion +sanitize-response command for outbound response filtering \n Integrates with Google Cloud authentication and global flags defined in gws-shared \n

Polish and refine agent prompt files against proven best practices. \n \n Requires a prompt file as input; will request one if not provided \n Preserves front matter, encoding, and markdown structure while improving clarity and organization \n Corrects spelling, grammar, and wording issues without altering the original intent \n Applies patterns from successful prompts to strengthen structure and effectiveness \n

Generates optimized prompts tailored to specific models and identity methods. Different models respond differently to prompts.