agent-skills
craigspaterson/agent-skills
Reusable skill definitions that teach AI agents consistent output formats, writing guidelines, and worked examples — works with Claude Code, GitHub Copilot, Cursor, and more.
Instalação
npx convez-ai craigspaterson/agent-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
agent-skills
A shared library of agent skill definitions. Skills teach your agent consistent output formats, writing guidelines, and worked examples so every team member gets the same quality of output without having to re-explain conventions in every conversation.
What Is a Skill?
A skill is a SKILL.md file that is loaded into the agent's context before a
task. It defines:
- Output format — the exact structure the agent should produce
- Writing guidelines — naming conventions, tone, and rules
- Acceptance criteria checklists — coverage prompts for common concerns
- Worked examples — minimal input → full output pairs
- Behavior rules — how the agent should handle edge cases and ambiguity
Available Skills
| Skill | Path | Purpose |
|---|---|---|
user-story | skills/user-story/SKILL.md | Jira-formatted User Stories with AC — for human-facing features, UI, and user journeys |
technical-story | skills/technical-story/SKILL.md | Jira-formatted Technical Stories with AC — for IaC, CI/CD, IAM, Cloud Run, Pub/Sub, and backend logic |
terraform-docs-inject | skills/terraform-docs-inject/SKILL.md | Regenerate and inject Terraform documentation into terraform/README.md using terraform-docs |
push-to-ado | skills/push-to-ado/SKILL.md | Upload a drafted User Story or Technical Story to Azure DevOps as a work item via upload.ps1 |
start-story | skills/start-story/SKILL.md | Activate an ADO work item (set state to Active, assign to self) and create + push the feature branch |
Install
Install skills in Claude Code, GitHub Copilot, Cursor, and more using skills.sh.
All Skills
npx skills add craigspaterson/agent-skills
Individual Skills
# User Story skill
npx skills add craigspaterson/agent-skills/skills/user-story
# Technical Story skill
npx skills add craigspaterson/agent-skills/skills/technical-story
# Terraform Docs Inject skill
npx skills add craigspaterson/agent-skills/skills/terraform-docs-inject
# Push to ADO skill
npx skills add craigspaterson/agent-skills/skills/push-to-ado
# Start Story skill
npx skills add craigspaterson/agent-skills/skills/start-story
Once installed, the agent activates a skill automatically when your request matches its description.
How to Use
Option 1 — Manual paste
- Open a new conversation with your agent.
- Copy the contents of the relevant
SKILL.mdfile. - Paste it at the start of your message, followed by your request.
Example:
[paste contents of skills/technical-story/SKILL.md]
Write a story for developing the cache-invalidation Cloud Run service in OpenTofu for the XYZ Platform.
Option 2 — Agent instructions file (recommended)
Most agent tools load an instructions file automatically at the start of every
session (AGENTS.md, CLAUDE.md, .cursorrules,
.github/copilot-instructions.md, etc.).
-
Clone this repo locally:
git clone git@github.com:<your-org>/agent-skills.git -
In your agent instructions file, reference the skill:
Before writing any user story, read skills/user-story/SKILL.md. Before writing any technical story, read skills/technical-story/SKILL.md. -
The agent will load the skill automatically when triggered.
Option 3 — Project knowledge (Team / Enterprise)
Many agent platforms support uploading files as shared project knowledge, making them available to all team members automatically.
- Create a project or workspace for your team.
- Upload the relevant
SKILL.mdfiles as project knowledge. - All team members will have the skills available automatically in every conversation.
Contributing
To add or update a skill:
- Branch from
main. - Add or edit the
SKILL.mdunderskills/<skill-name>/. - Update the table in this README.
- Open a PR for review — skills affect output quality for the whole team, so a second pair of eyes is worthwhile.
Skill files follow a standard frontmatter block at the top:
---
name: skill-name
description: One-sentence trigger description used by the agent to decide when to load this skill.
---
Questions
Reach out to the team or raise a GitHub issue in this repo.