agent-skills

agent
Guvenlik Denetimi
Uyari
Health Uyari
  • No license — Repository has no license file
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 7 GitHub stars
Code Gecti
  • Code scan — Scanned 5 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Collection of agent skills that I think are useful

README.md

Agent Skills

A collection of skills for AI coding agents. Skills are packaged instructions that extend agent capabilities, from commit hygiene to full design system references.

Skills follow the Agent Skills format.

Skills at a Glance

Skill What it does Trigger phrases
conventional-commits Commit messages following the Conventional Commits spec "Write a commit message", "Commit my changes"
conventional-comments Structured code review comments "Write review comments", "Give PR feedback"
pr-message Concise, high-signal PR descriptions "Write a PR description"
pr-review Analyze a PR for effective review "Review this PR", "Explain this pull request"
challenge-me Blunt technical advisor mode "Challenge my design", "Pressure-test this"
humanizer Remove signs of AI writing from text "Humanize this text", "Make it sound less AI"
apple-human-interface-guidelines Apple HIG reference across all platforms "Review my UI for HIG compliance"
google-adk Google Agent Development Kit reference for building AI agents "Build an ADK agent", "Deploy my ADK agent to GCP"

Developer Workflow

conventional-commits

Generate commit messages following the Conventional Commits v1.0.0 specification with Angular convention types.

Use when: creating git commits, writing commit messages, or enforcing consistent commit conventions.

Types: feat, fix, build, chore, ci, docs, style, refactor, perf, test

conventional-comments

Format review comments following the Conventional Comments standard. Provides labels, decorations, and communication best practices for clear, actionable feedback.

Use when: writing code review comments, giving PR feedback, or any written review process.

Labels: praise, nitpick, suggestion, issue, todo, question, thought, chore, note (plus optional typo, polish, quibble)

Decorations: (non-blocking), (blocking), (if-minor), and custom (e.g., security, ux)

pr-message

Write concise, high-signal GitHub pull request descriptions that explain intent, impact, and risk without duplicating information already visible in GitHub.

Use when: opening a pull request and you need a well-structured summary.

Sections: Summary, What Changed, Testing, Screenshots (UI only), Risk / Rollout, Notes for Reviewers

pr-review

Analyze and explain a pull request to help review it effectively. Summarizes what changed, why, how the pieces fit together, and what risks deserve attention.

Use when: preparing to review someone else's code or trying to understand a PR quickly.

Analysis areas: Understanding, Risk & Correctness, Architecture & Consistency, Review Guidance, UI / Functional Changes

challenge-me

Direct, no-comfort technical advisor mode. Challenges ideas, questions assumptions, and surfaces blind spots.

Use when: evaluating tradeoffs, scoping work, or pressure-testing an architecture decision.

Focus areas:

  • Architecture & Design — pressure-test boundaries, ask what can fail, call out unnecessary
  • Counting & Estimation — break "quick/easy" into concrete tasks, surface hidden work
    complexity

Writing

humanizer

Detect and remove signs of AI-generated writing from text. Based on documented, research-backed patterns from Wikipedia's Signs of AI Writing.

Use when: writing, editing, or rewriting text to sound more natural and avoid AI detection patterns.

What it catches:

Category Examples
AI vocabulary "delve", "tapestry", "vibrant", "pivotal", "foster", "underscore"
Inflated significance "serves as a testament to", "marking a pivotal moment"
Promotional puffery "nestled in the heart of", "boasts a diverse array"
Structural tells "not just X, but also Y", rule-of-three, "despite challenges"
Grammar quirks "serves as" instead of "is", synonym cycling, dangling "-ing" phrases
Style artifacts title case headings, bold-colon lists, em dash overuse

Design & UI

apple-human-interface-guidelines

Comprehensive Apple Human Interface Guidelines reference covering all Apple platforms. Contains 150+ reference documents organized by category.

Use when: designing or reviewing UI for iOS, iPadOS, macOS, tvOS, watchOS, or visionOS — or applying Apple's design principles to any app.

Platforms: iOS, iPadOS, macOS, tvOS, watchOS, visionOS

Coverage:

Category What's included
Getting Started Platform-specific design fundamentals and constraints
Foundations Accessibility, color, dark mode, typography, layout, motion, privacy, branding
Patterns Data entry, search, feedback, loading, onboarding, modality, navigation, media
Components Buttons, menus, lists, tabs, sheets, alerts, pickers, toggles, charts, and more
Inputs Touch, Apple Pencil, keyboard, game controllers, Digital Crown, gaze (visionOS)
Technologies Apple Pay, Sign in with Apple, HealthKit, ARKit, CarPlay, SharePlay, Generative AI

Common tasks quick reference:

Task References loaded
Building a login screen Privacy, Sign in with Apple, Managing Accounts
Designing a settings screen Settings patterns, Toggles
Choosing colors / dark mode Color guidelines, Dark Mode
Implementing navigation Tab bars, Sidebars, Split views
Accessibility audit Accessibility foundations, VoiceOver
Designing a form Entering data, Text fields, Pickers
Building for visionOS visionOS fundamentals, Spatial layout, Immersive experiences

Many foundations (accessibility, color, typography, motion, privacy, writing) are applicable beyond Apple platforms to web and cross-platform development.


AI Agent Development

google-adk

Google Agent Development Kit reference for building AI agents in Python, TypeScript, Go, and Java. Contains 90+ reference files organized by topic. The 10 most code-heavy files are split into per-language sidecars so you load only what you need.

Use when: building or configuring ADK agents, wiring tools, managing sessions and state, setting up multi-agent pipelines, streaming with the Gemini Live API, deploying to GCP, or evaluating agent behavior.

Languages: Python (primary), TypeScript, Go, Java

Coverage:

Category What's included
Agents LlmAgent, SequentialAgent, LoopAgent, ParallelAgent, custom BaseAgent, multi-agent routing
Tools Function tools, MCP, OpenAPI, tool authentication, action confirmations
Sessions & Memory Session service, state prefixes, MemoryService, cross-session recall
Callbacks Before/after model, tool, and agent callbacks
Streaming Gemini Live API, LiveRequestQueue, run_live(), audio/video/images
Deployment Vertex AI Agent Engine, Cloud Run, GKE
Context Context caching, context compression (EventsCompactionConfig)
Evaluation Criteria, test writing, user simulation
A2A Protocol Agent-to-Agent communication
Observability Logging, Cloud Trace

Installation

npx skills add milistu/agent-skills

Usage

Skills are automatically available once installed. The agent uses them when relevant tasks are detected.

Write a commit message for my changes
Write review comments for this PR
Write the PR description
Review this pull request
Challenge my design for this feature
Humanize this text — it sounds too AI-generated
Review my UI against Apple HIG
How do I add OAuth2 authentication to my ADK tool?

Skill Structure

Each skill contains:

  • SKILL.md — Instructions for the agent
  • reference/ — Supporting documentation (optional)

License

MIT

Yorumlar (0)

Sonuc bulunamadi