Skip to main content

create-skill

Create Agent Skills for ORG2. Use when the user wants to create, write, or author a new skill, capture a workflow as a skill, or asks about SKILL.md format, skill structure, or best practices.

Quellinformationen

Repository
org2AI/ORG2
Letzte Quellaktivität
14. September 2026 um 14:43
Erkannte Sprache von SKILL.md
Englisch
Sterne
2.613
Forks
128

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

SKILL.md wird angezeigt

SKILL.md
Quellanweisungen · Schreibgeschützte Vorschau
name
create-skill
description
Create Agent Skills for ORG2. Use when the user wants to create, write, or author a new skill, capture a workflow as a skill, or asks about SKILL.md format, skill structure, or best practices.
# Creating Skills in ORG2 Skills are markdown files that teach the agent how to perform specific tasks. They appear in the `/` slash menu and are included in the agent's system prompt when relevant. ## Before You Begin Gather from the user: 1. **Purpose**: What task or workflow should this skill help with? 2. **Location**: Global (`~/.orgii/skills/`) or Project (`<repo>/.orgii/skills/`)? 3. **Trigger scenarios**: When should the agent use this skill? 4. **Domain knowledge**: What specialized context does the agent need? If you have conversation context, infer the skill from what was discussed. ## Storage Locations | Type | Path | Scope | | ------- | ---------------------------------- | ----------------------------- | | Global | `~/.orgii/skills/skill-name/` | Available across all projects | | Project | `<repo>/.orgii/skills/skill-name/` | Shared with the repository | These two scopes match the `SKILL_SCOPE` constant in `src/types/extensions/types.ts` and the picker labels in `SkillEditorPanel.tsx`. Skills do not have a separate "Personal" scope the way Rules do — there is only one per-user location, and it is called **Global**. ## Skill File Structure ``` skill-name/ ├── SKILL.md # Required — main instructions ├── reference.md # Optional — detailed docs / reference tables └── examples.md # Optional — usage examples ``` ### SKILL.md Format ```markdown --- name: your-skill-name description: Brief description of what this skill does and when to use it. --- # Skill Name ## Instructions Clear, step-by-step guidance for the agent. ## Examples Concrete examples. ``` ### Frontmatter Fields | Field | Requirements | | ------------- | --------------------------------------------------------- | | `name` | Max 64 chars, lowercase letters/numbers/hyphens only | | `description` | Max 1024 chars. Include WHAT it does AND WHEN to trigger. | ## Writing a Good Description The description is critical — the agent uses it to decide when to apply the skill: ```yaml # Good — specific, includes trigger terms description: Deploy Rust services to AWS Lambda. Use when building or deploying Lambda functions, setting up SAM templates, or packaging Rust for serverless. # Bad — too vague description: Helps with deployment ``` **Rules:** 1. Write in third person ("Use when..." not "I help with...") 2. Include both WHAT and WHEN 3. List concrete trigger terms (tool names, file types, workflows) ## Authoring Principles - **Concise**: Only include context the agent doesn't already have - **Under 500 lines**: Keep SKILL.md focused; put details in `reference.md` - **Progressive disclosure**: Link to reference files rather than inlining everything - **Actionable**: Steps should be executable, not aspirational ## Creation Workflow 1. **Clarify** — gather purpose, location, triggers, constraints from user 2. **Design** — draft name, description, section outline 3. **Implement** — create the directory and write SKILL.md 4. **Verify** — check description quality, line count, name format ## Example Skill ```markdown --- name: code-review description: Review code changes for quality, security, and maintainability. Use when reviewing pull requests, diffs, or checking code before commit. --- # Code Review ## Review Order 1. Correctness — does the code do what it claims? 2. Security — are there injection, auth, or data-leak risks? 3. Readability — would a new team member understand this? 4. Tests — are edge cases covered? ## Feedback Labels - **Critical** — must fix before merge - **Suggestion** — worth improving - **Nit** — minor style preference ## Reference See [review-checklist.md](review-checklist.md) for the full checklist. ``` ## Checklist - [ ] Directory created at the correct location - [ ] `name` is lowercase-hyphenated, max 64 chars - [ ] `description` includes trigger terms and WHEN clause - [ ] SKILL.md is under 500 lines - [ ] Heavy reference material moved to `reference.md`
Auf GitHub ansehen