Skip to content

inkeep/open-knowledge / packages/server/assets/skills/write-skill

open-knowledge-write-skill

Verified

Use when the user wants to create, author, write, or design a new Agent Skill (a SKILL.md) — for OpenKnowledge or for their editors — including requests like 'help me write a skill', 'make a skill that…', 'turn this workflow into a skill', or improving an existing skill's triggering and discipline. Also use when capturing reusable agent guidance that should live as an installable skill rather than a one-off prompt. Covers choosing scope (project vs global), authoring inside a plugin or skills-distribution repo (write in the repo's own layout, never install), the SKILL.md frontmatter contract, progressive-disclosure structure, evaluating the skill, and installing it into the user's editors.

FileKindSize
SKILL.mdskill md14.7 KB
references/description-optimization.mdmarkdown1.7 KB
references/pressure-testing.mdmarkdown1.9 KB

references/description-optimization.md

View on GitHub
# Optimizing a skill's `description`

The `description` is the **entire** signal the agent uses to decide whether to
load a skill. The body doesn't exist until the description fires. Optimize it
last, deliberately.

## Rules

1. **Triggers, not a summary.** Describe WHEN to reach for the skill, in the
   user's words. Do NOT recap what the body does — empirically, summarizing the
   workflow in the description makes the model follow the description and skip
   the body (the classic "did one review instead of the two the body required").
2. **Be concrete and slightly pushy.** Under-triggering (the skill never loads)
   is the common failure. Name the situations, the verbs, and the actual
   phrasings a user would type. List several; don't rely on one abstract clause.
3. **Stay inside the contract.** ≤1024 characters, no XML/angle-bracket tags, no
   `version` field. A tag like `<thing>` breaks the loader.
4. **Disambiguate from neighbors.** If a sibling skill is close, add a line on
   when to use THIS one vs the other.

## A lightweight near-miss check

1. Write ~10 phrasings that SHOULD trigger the skill and ~5 adjacent ones that
   should NOT (they belong to a different skill or no skill).
2. Read the description cold and predict, for each phrasing, whether it would
   fire. Every should-fire that you'd miss is a gap; every should-not that fires
   is over-reach.
3. Edit the description to close the gaps without grabbing the should-nots.
   Prefer adding concrete trigger phrasings over broadening vague ones.

## Shape

Lead with the dominant trigger ("Use when the user wants to…"), then enumerate
concrete situations and phrasings, then any disambiguation. Keep it scannable.
brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).Stars: 318kUpdated Claude CodeCodexVerifiedInstall
brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).Stars: 196kUpdated Claude CodeCodexVerifiedInstall