Skills

A collection of AI agent skills — reusable procedural knowledge packages that extend AI assistants like GitHub Copilot, Claude, and others. Each skill encodes domain expertise, conventions, and workflows that agents can follow when performing specialized tasks.

Skills

Skill	Description
agentic-eval	Patterns for evaluating and improving AI agent outputs — self-critique loops, evaluator-optimizer pipelines, rubric-based evaluation
azure-cost-analysis	Analyze Azure costs via the CostManagement REST API — query construction, breakdowns by resource/meter/service, anomaly detection
azure-naming-convention	Generate compliant Azure resource names following Microsoft Cloud Adoption Framework (CAF) best practices
azure-verified-modules	Develop and consume certified Azure Verified Modules (AVM) for Terraform — VNet injection, subnet delegation, NSG rules
cli-creator	Create CLI tools following clig.dev best practices — supports Python (Click/Typer), Node.js (Commander), Go (Cobra), Rust (Clap), and Bash
eval-audit	Audit LLM eval pipelines to surface problems — missing error analysis, unvalidated judges, vanity metrics
fastapi	Build FastAPI applications with uv project setup, Pydantic models, dependency injection, async endpoints, SSE streaming
generate-synthetic-data	Create diverse synthetic test inputs for LLM pipeline evaluation using dimension-based tuple generation
java-best-practices	Modern Java best practices (JDK 8–25) — records, sealed classes, pattern matching, virtual threads, Spring Boot, JUnit 5
microsoft-agent-framework	Build AI agents using the Microsoft Agent Framework Python SDK 1.0.0 — OpenAI, Azure OpenAI, Foundry, Anthropic providers
python-best-practices	Modern Python best practices — PEP 8, type hints, pytest, dataframe workflows (Pandas/Polars/DuckDB), Python data model patterns
skill-creator	Create and update Agent Skills following the agentskills.io specification
skillshare	Manage and sync AI CLI skills across 50+ tools from a single source
terraform-style-guide	Generate Terraform HCL following HashiCorp's official style conventions and best practices
write-judge-prompt	Design LLM-as-Judge evaluators for subjective criteria that code-based checks cannot handle

Structure

Each skill follows a consistent layout:

skill-name/
├── SKILL.md              # Main skill file with YAML frontmatter and instructions
├── references/           # Supporting reference material loaded on demand
│   ├── topic-a.md
│   └── topic-b.md
└── evals/                # Optional evaluation tasks for the skill

• SKILL.md — The entry point. Contains the skill name, description (used for matching), and the full procedural instructions the agent follows.
• references/ — Supplementary docs that the skill loads as needed to keep the main file focused.
• evals/ — Task definitions for testing skill quality (see eval/ at the repo root).

Usage

With Skillshare

Install skills to any supported AI tool using skillshare:

skillshare install <skill-name>
skillshare sync

Manual

Copy or symlink a skill directory into your AI tool's skill/prompt configuration folder. For example, with GitHub Copilot in VS Code, place skills under ~/.copilot/skills/.

Evals

The eval/ directory contains evaluation scripts that test skill quality:

eval/
├── eval_azure_verified_modules.py
├── eval_cli_creator.py
├── eval_java_best_practices.py
├── eval_skill_creator.py
└── eval_terraform_style_guide.py