Guides the creation of new SKILL.md files with discovery-focused declarations.
npx skills add https://github.com/Eyhenij/rt-tools --skill write-a-skill-eyhenij
Please help me install this Agent Skill. Skill: write-a-skill Source: https://github.com/Eyhenij/rt-tools/tree/main/.claude/skills/write-a-skill Command: npx skills add https://github.com/Eyhenij/rt-tools --skill write-a-skill-eyhenij
description frontmatter field — the only text the agent sees when deciding whether to load a skill — including the "Load when…" pattern and naming the neighbour not to load. - Structure Rules: Defines the file layout (SKILL.md first, optional extra files, scripts/ for deterministic operations) and when to split a growing skill. - Quality Checklist: Provides a pre-handover checklist covering staleness, duplication, length, and link depth, plus guidance on making a skill discoverable via gate maps or other skills. - Use Case: When asked to create a new skill for a showcase, generator, third-party service, or team technique, follow this Skill to draft the SKILL.md, write a loadable declaration, and validate it before handover. ## Quick Start Ask the agent to create a new skill for your task and have it follow the write-a-skill process for the declaration and structure.High-intent search queries and answers about installing and using this skill.
Write the description in third person, up to 1024 characters: first sentence states what the skill does, second says "Load when…", third names what is not covered and which neighbouring skill handles it. The description is the only text the agent sees when choosing skills.
A skill needs a SKILL.md with YAML frontmatter containing name and description, followed by a title and sections starting with "When to use". Add extra files only when SKILL.md stops being readable in full, and a scripts/ directory only for deterministic operations.
Split when the file exceeds roughly a hundred lines, holds two distinct subjects where one is rarely needed, or ready-made code outweighs its explanation. Do not split for aesthetics, since each extra file adds a reading hop.
Skills go unused when the description lacks a clear "Load when" trigger or fails to distinguish itself from neighbours. A skill must also be referenced in the tree's gate map or called by name from another skill or role.
Avoid retelling tool documentation the agent already knows, duplicating neighbouring skills, and anything that goes stale like versions, numbers, or dates. Rules and patterns governed by a law belong to the spec-driven-rule pattern instead.
Mateus2411Check and invoke applicable superpowers skills before any assistant response.
agent-ecosystemTests discovery of a nested SKILL.md via a parent references directory.
hoanghn61Resolve contradictory technology signals and produce a structured resolution report.
udecodeDefine or update editor behavior specs using the research-to-law workflow.