ai-engineering
AI Engineering
Overview
Build effective agentic systems using proven patterns. Start simple, add complexity only when needed.
For specialized prompt design guidance (techniques, patterns, examples for agentic systems), see the prompt-engineering skill.
Core Principle
Find the simplest solution first. Agentic systems trade latency and cost for better task performance. Only increase complexity when simpler solutions fall short.
- Start with optimized single LLM calls (retrieval, in-context examples)
- Add workflows for predictable, multi-step tasks
- Use agents when flexibility and autonomous decision-making are required
When to Build an Agent
Before committing to an agent, validate that your use case truly requires agentic capabilities. Consider alternatives first—deterministic solutions are simpler, faster, and more reliable.
More from mguinada/agent-skills
refactor
TDD-based code refactoring preserving behavior through tests. Use Red-Green-Refactor cycles to apply refactoring patterns one test-verified change at a time. **TRIGGERS**: 'clean up code', 'make code simpler', 'reduce complexity', 'refactor this', 'apply DRY', 'extract method', 'remove duplication'. **DISTINCT FROM**: Adding features (use /tdd) or fixing bugs. **PROACTIVE**: Auto-invoke when test-covered code has complexity (functions >50 lines, high cyclomatic complexity, duplication).
16git-commit
Generate concise, descriptive git commit messages following best practices. Use when creating git commits from staged changes, crafting commit messages, or reviewing commit message quality. Use when the user says /commit or asks to create a git commit. **PROACTIVE ACTIVATION**: Auto-invoke when staged changes detected or user asks to commit/save work. **DETECTION**: Run git status - if staged changes exist, offer to commit. User says \"commit\", \"save\", \"done with feature\". **USE CASES**: Staged changes detected, work completed, user wants to save progress.
12tdd
Guide Test-Driven Development workflow (Red-Green-Refactor) for new features, bug fixes, and refactoring. Supports both Python (pytest) and Ruby (RSpec). Use when writing tests, implementing features, or following TDD methodology. **PROACTIVE ACTIVATION**: Auto-invoke when implementing features or fixing bugs in projects with test infrastructure. **DETECTION**: Check for tests/ directory, pytest.ini, pyproject.toml with pytest config, spec/ directory, .rspec file, or *_spec.rb files. **USE CASES**: Writing production code, fixing bugs, adding features, legacy code characterization.
11create-pr
Creates GitHub pull requests with properly formatted titles, a body matching the project's PR template, and appropriate type/scope labels. Automatically creates labels if they don't exist. Use when creating PRs, submitting changes for review, or when the user says /pr or asks to create a pull request. **PROACTIVE ACTIVATION**: Auto-invoke when a branch has commits ahead of main and the user signals the work is ready. **DETECTION**: Run git log origin/main..HEAD - if commits exist and user signals readiness, offer to open a PR. User says \"open a PR\", \"ready for review\", \"this is done\", \"let's merge\", \"submit this\". **USE CASES**: Feature or fix complete, user finished a series of commits and mentions review or merging.
11prompt-engineering
Creates system prompts, writes tool descriptions, and structures agent instructions for agentic systems. Use when the user asks to create, generate, or design prompts for AI agents, especially for tool-using agents, planning agents, or autonomous systems. **PROACTIVE ACTIVATION**: Auto-invoke when designing prompts for agents, tools, or agentic workflows in AI projects. **DETECTION**: Check for agent/tool-related code, prompt files, or user mentions of \"prompt\", \"agent\", \"LLM\". **USE CASES**: Designing system prompts, tool descriptions, agent instructions, prompt optimization, reducing hallucinations.
10copilot-sdk
Build agentic applications with GitHub Copilot SDK. Use when embedding AI agents in apps, creating custom tools, implementing streaming responses, managing sessions, connecting to MCP servers, or creating custom agents. Triggers on Copilot SDK, GitHub SDK, agentic app, embed Copilot, programmable agent, MCP server, custom agent. **PROACTIVE ACTIVATION**: Auto-invoke when building agentic applications or integrating Copilot SDK. **DETECTION**: Check for @github/copilot-sdk imports, copilot dependencies in package.json/pyproject.toml/go.mod. **USE CASES**: Embedding agents in apps, creating custom tools, implementing streaming, managing sessions, connecting to MCP servers.
9