A portfolio of 17 Agent Skills covering the full software development lifecycle for Python (FastAPI) + React/TypeScript projects. Built on the Agent Skills open standard — works with Claude Code, Cursor, GitHub Copilot, Codex, Windsurf, and other compatible tools.

npx skills add hieutrtr/ai1-skills

Try these prompts after installing:

"Plan the implementation for adding user authentication"
"Create a FastAPI endpoint for user registration"
"Review this code for security vulnerabilities"

See Skill Composition Guide for how skills relate to each other.

flowchart LR
    subgraph Planning
        PP["project-planner"] --> TD["task-decomposition"]
    end

    subgraph Architecture
        SA["system-architecture"]
        AD["api-design-patterns"]
    end

    subgraph Implementation
        BE["python-backend-expert"]
        FP["fastapi-patterns"]
        FE["react-frontend-expert"]
    end

    subgraph Testing
        TDD["tdd-workflow"]
        PY["pytest-patterns"]
        RT["react-testing-patterns"]
        E2E["e2e-testing"]
    end

    subgraph "Code Review"
        SEC["code-review-security"]
        PMC["pre-merge-checklist"]
    end

    subgraph Deployment
        DOC["docker-best-practices"]
        DEP["deployment-pipeline"]
    end

    subgraph Operations
        MON["monitoring-setup"]
        IR["incident-response"]
    end

    Planning --> Architecture --> Implementation --> Testing --> Code_Review["Code Review"] --> Deployment --> Operations

Install the entire skill portfolio with one command:

npx skills add hieutrtr/ai1-skills

This downloads all 17 skills into your project's .claude/skills/ directory.

To install a single skill:

npx skills add hieutrtr/ai1-skills --skill "python-backend-expert"

Clone the repository directly into your project:

# Project-scoped (this project only)
git clone https://github.com/hieutrtr/ai1-skills.git .claude/skills-repo
cp -r .claude/skills-repo/skills/* .claude/skills/
rm -rf .claude/skills-repo

# Personal scope (all your projects)
git clone https://github.com/hieutrtr/ai1-skills.git ~/.claude/skills-repo
cp -r ~/.claude/skills-repo/skills/* ~/.claude/skills/
rm -rf ~/.claude/skills-repo

Download individual skill directories from github.com/hieutrtr/ai1-skills and place them in one of these locations:

Open Claude Code in your project and ask:

What skills are available?

Claude should list all 17 skills. You can also check context usage:

/context

Expected startup cost: 17 skills x ~100 tokens = ~1,700 tokens for Level 1 metadata.

Skills use a progressive disclosure model with three levels:

1. Level 1 — Metadata (~100 tokens per skill): Skill descriptions are loaded at startup so Claude knows what's available.
2. Level 2 — Full skill: When Claude determines a skill is relevant to your request, it loads the complete SKILL.md into context.
3. Level 3 — References: Supporting files (references/, scripts/) are loaded on-demand only when the active skill references them.

You don't need to do anything special. Claude activates the right skill based on your request.

Just ask Claude naturally. The skill descriptions contain phase-specific keywords that trigger the correct skill:

# Activates project-planner
"Plan the implementation for adding user authentication"

# Activates python-backend-expert
"Create a new endpoint for user registration"

# Activates pytest-patterns
"Write tests for the user service"

# Activates code-review-security
"Review this code for security vulnerabilities"

# Activates deployment-pipeline
"Set up the CI/CD pipeline for this project"

Invoke any skill directly with its name as a slash command:

/project-planner Add a payment processing module
/python-backend-expert Create CRUD endpoints for orders
/code-review-security Review the auth module
/pre-merge-checklist Run all quality checks
/tdd-workflow Implement the search feature using TDD

Skills compose across SDLC phases — each phase produces artifacts consumed by the next.

For detailed composition rules, activation boundaries, and workflow diagrams, see:

• Skill Composition Guide — how skills activate, relate, and when not to use one
• Greenfield Guide — full workflow for new projects (phases 1-8 in order)
• Brownfield Guide — incremental adoption for existing projects

Skills become more powerful when paired with MCP servers that provide live data access:

Each skill is a directory with this structure:

skill-name/
├── SKILL.md              # Main instructions (required, <500 lines)
├── references/           # Detailed docs, templates, examples (on-demand)
│   ├── template.md
│   └── patterns.md
└── scripts/              # Executable validation/automation scripts
    └── check.sh

Every SKILL.md has YAML frontmatter and Markdown content:

---
name: python-backend-expert
description: >-
  Python backend patterns for FastAPI with SQLAlchemy 2.0, Pydantic v2,
  and async patterns. Use during implementation when creating endpoints,
  models, or services. Does NOT cover testing (use pytest-patterns).
license: MIT
compatibility: 'Python 3.12+, FastAPI 0.115+, SQLAlchemy 2.0+, Pydantic v2'
metadata:
  author: platform-team
  version: '1.0.0'
  sdlc-phase: implementation
allowed-tools: Read Edit Write Bash(python:*) Bash(pip:*) Bash(alembic:*)
context: fork
---

# Python Backend Expert

## When to Use
...

## Instructions
...

## Examples
...

## Edge Cases
...

Key frontmatter fields:

• description — Controls when Claude activates the skill. Include phase keywords and negative keywords ("Does NOT cover...")
• allowed-tools — Restricts what tools Claude can use. Planning skills get read-only access; implementation skills get write access.
• context: fork — Runs the skill in an isolated subagent context

These skills encode conventions for:

With all 17 skills installed, the context impact is minimal:

This is under 8% of a 200K context window.

To customize a skill for your project, copy it to your project's .claude/skills/ and edit. Project-scoped skills take precedence over personal-scoped ones.

Edit the relevant SKILL.md to add your team's conventions. For example, add your database naming conventions to python-backend-expert, or your component library patterns to react-frontend-expert.

Add files to references/ for large reference material. Reference them from SKILL.md:

See [API catalog](references/api-catalog.md) for endpoint documentation.

Claude loads these only when the skill is active and references them.

These skills follow the agentskills.io core standard and work with:

• Claude Code
• Cursor
• GitHub Copilot (VS Code)
• OpenAI Codex
• Windsurf
• Other agents supporting the Agent Skills standard

MIT