https://github.com/yu-iskw/coding-agent-skills
A curated collection of executable Agent Skills and SOPs for AI coding agents like Cursor, Claude Code, and Gemini CLI.
https://github.com/yu-iskw/coding-agent-skills
agent-skill agent-skills claude-code codex codex-cli gemini-cli
Last synced: 14 days ago
JSON representation
A curated collection of executable Agent Skills and SOPs for AI coding agents like Cursor, Claude Code, and Gemini CLI.
- Host: GitHub
- URL: https://github.com/yu-iskw/coding-agent-skills
- Owner: yu-iskw
- Created: 2026-01-31T15:54:29.000Z (18 days ago)
- Default Branch: main
- Last Pushed: 2026-02-02T06:10:40.000Z (17 days ago)
- Last Synced: 2026-02-02T18:05:09.139Z (16 days ago)
- Topics: agent-skill, agent-skills, claude-code, codex, codex-cli, gemini-cli
- Language: Shell
- Homepage:
- Size: 38.1 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- Codeowners: .github/CODEOWNERS
Awesome Lists containing this project
README
# Coding Agent Skills: The Orchestration Framework for Industrial-Grade AI Workflows
Standard coding agents are powerful but often hit a **"Complexity Ceiling"** in long-horizon tasks. As conversations grow, context decays, hallucinations increase, and reliability drops.
**Coding Agent Skills** transform your repository from a collection of code into an **Agent-Ready Environment**. By codifying workflows into deterministic, reusable skills, you enable agents to work with higher autonomy, precision, and parallelism.
## Why Coding Agent Skills?
Standard agents operate in a single-session vacuum. Coding Agent Skills provide:
- **Orchestration**: Decouple planning from execution for complex tasks.
- **Reliability**: Replace vague prompts with structured, executable SOPs.
- **Efficiency**: Parallelize work that would take a single agent hours to complete.
## Key Advantages for Advanced Users
1. **Multi-Agent Parallelism**: Break monolithic tasks into sub-tasks and execute them simultaneously across multiple agent instances.
2. **Deterministic SOPs**: Standardize complex workflows (e.g., security audits, library upgrades) as executable SOPs that ensure zero steps are skipped.
3. **Context Hygiene**: Use progressive disclosure to feed agents only the data they need, maintaining high reasoning quality.
4. **Autonomous Fixer Loops**: Encapsulate `Test -> Analyze -> Fix` loops so agents self-correct without human intervention.
5. **Platform Portability**: Write skills once; run them across **Cursor**, **Claude Code**, **Gemini CLI**, and **Codex**.
6. **Safety Governance**: Define granular permission modes per skill to manage risk in autonomous executions.
7. **Institutional Memory**: Codify "Tribal Knowledge" into skills so every agent (and human) operates at an expert level.
8. **Sub-Agent Specialization**: Spin up micro-personas (e.g., "SQL Optimizer") for domain-specific precision.
9. **Headless CI/CD**: Trigger skills in pipelines for automated PR validation and structured reporting.
10. **Low-Code Customization**: Add new capabilities by simply adding a `SKILL.md` file—no extension development required.
## Building an Agent-Ready Repository
An Agent-Ready repository isn't just about good code; it's about **explicit intent**. By providing `SKILL.md` files, you give AI agents a clear "map" of how to interact with your codebase, what tools to use, and how to validate their own work.
## Available Skills
This repository includes the following coding agent skills:
- **[orchestrating-parallel-tasks](skills/orchestrating-parallel-tasks/)**: Decomposes complex or large-scale requirements into discrete, mutually exclusive sub-tasks that can be executed in parallel by multiple agents. Use when a task is described as an "epic", spans multiple architectural layers, or would benefit from concurrent development to reduce time-to-completion.
- **[claude-code-cli](skills/claude-code-cli/)**: Executes tasks using the Claude Code CLI (`claude`). Automatically determines the safest permission mode (plan, acceptEdits, or bypassPermissions) based on the task type.
- **[codex-cli](skills/codex-cli/)**: Executes tasks using the Codex CLI (`codex`). Automatically determines the least privilege required (Read-Only, Editor, or Autonomous) based on the user's request and handles security approvals.
- **[gemini-cli](skills/gemini-cli/)**: Executes tasks using the Gemini CLI (`gemini`). Automatically determines the safest approval mode (plan, auto_edit, or yolo) based on the task type.