Prompt Writing

token-budget-advisor is a routing skill that helps choose response depth before answering. Use it to control length, detail, or token spend in workflow automation and chat flows. This token-budget-advisor guide covers trigger rules, when not to use it, and how to apply it from skills/token-budget-advisor.

strategic-compact helps you decide when to manually run /compact so Claude sessions stay coherent at task boundaries. It supports long, multi-phase work, especially research, planning, implementation, and testing, and is useful for workflow automation when you want compaction to happen at logical milestones instead of arbitrary auto-compaction.

cost-aware-llm-pipeline helps you build LLM workflows that control API spend with model routing, immutable cost tracking, retry handling, and prompt caching. Ideal for batch jobs, document pipelines, and Workflow Automation where output volume and quality tradeoffs need clear rules.

caveman is an ultra-compressed communication skill for concise technical replies, summaries, and debugging help. It keeps technical meaning intact, drops filler, and stays active after trigger until you say stop caveman or normal mode. Best for Technical Writing, code review notes, incident updates, and faster back-and-forth when you already know the topic.

p9 is a tech-lead style skill for Agent Orchestration that writes task prompts, coordinates P8 agents, and avoids direct coding. Use it to split project goals into scoped, executable prompts with roles, constraints, dependencies, and acceptance criteria.

ask-questions-if-underspecified helps agents pause on unclear requests, ask the minimum clarifying questions, and avoid wrong work. This ask-questions-if-underspecified skill is useful for Skill Authoring, coding tasks, and any workflow where objective, scope, or constraints are missing.

huashu-nuwa is a Skill Authoring skill that turns a person, topic, or vague need into a reusable perspective skill. It guides install and usage through `SKILL.md`, `references/extraction-framework.md`, `references/skill-template.md`, and example skills for distillation, routing, and quality review.

llm-patterns helps you design AI-first application logic where LLMs handle reasoning, extraction, and generation while code handles validation, routing, and error handling. Use the llm-patterns skill for clearer prompt structure, testable LLM workflows, and practical guidance for Skill Authoring.

writing-fragments is a focused interview skill for capturing claims, vignettes, sharp sentences, and half-formed ideas as raw material for a future article. It fits Knowledge Capture and early-stage drafting when you want to collect fragments first and defer structure until later. Use writing-fragments when you need a practical writing-fragments guide instead of a normal outline prompt.

moyu-lite is a lightweight guardrail skill for code editing that keeps changes tightly scoped to the user’s exact request. It avoids extra refactors, new abstractions, broad rewrites, and side changes, making moyu-lite useful when you need minimal diffs, careful scope control, and a clear moyu-lite guide for restrained editing.

moyu-fr is a code-editing guardrail skill for small, surgical changes. It helps agents avoid overengineering, keep diffs minimal, and focus on the exact edit requested. Use moyu-fr when you want a simple patch, not a rewrite, and when tight scope control matters more than expansion.

moyu-en is a restraint-first skill for code editing. It helps keep changes surgical, avoid over-engineering, and limit diffs to exactly what the user asked for. Use moyu-en when you need precise, small patches, narrow scope control, and a better guide for minimal code updates.

prd-generator turns a rough product idea into an AI-ready Product Requirements Document. It asks clarifying questions, follows a fixed template, and helps founders, product leads, and Skill Authoring workflows produce clearer specs for downstream AI coding tools. Use prd-generator when you need structured requirements, metrics, constraints, and implementation-ready context.

context-engineering-advisor helps you diagnose context stuffing vs. context engineering, then tighten boundaries, retrieval, and workflow order for more reliable AI output.

Use openai-docs for Technical Writing, OpenAI API and product questions, model selection, migration checks, and prompt-upgrade guidance. It prioritizes official OpenAI docs via the Developer Docs MCP server, with bundled references as fallback context only when needed.

pua-ja is a Japanese-language escalation skill that pushes stalled agents to investigate harder, use tools before asking users, and verify results after repeated failures. Best for teams that want a trigger-based behavior layer for debugging, research, writing, and pua-ja for Context Engineering.