Coding Agent Tools & Workflows
This skill guides the setup and usage of the modern "AI Pair Programmer" stack.
How to select AI Coding Tools
Choose tools based on interaction depth.
See CLI Guide for detailed comparisons.
| Tool | Type | Best Use Case |
|---|---|---|
| Claude Code / Aider | CLI Agent | Full-file refactoring, multi-file edits, "Apply this feature". |
| GitHub Copilot | IDE Ghostwriter | Autocomplete, boilerplate generation, quick unit tests. |
| Cursor | AI-Native IDE | "Chat with codebase", semantic search, diff review. |
| OpenCode | Open Interpreter | Local LLM execution, privacy-focused tasks. |
How to workflow with AI Agents (The "Coding Vibe")
- Spec-First: Don't write code. Write a Markdown spec (
spec.md) and ask the agent to "Implement this spec". - TDD Loop:
- Ask agent: "Generate tests for feature X based on spec."
- Run tests (Fail).
- Ask agent: "Write implementation to pass tests."
- Run tests (Pass).
- Reviewer Mode: Before committing, pipe
git diffto the agent: "Review this diff for security issues and potential bugs."
How to integrate AI into CI/CD
- PR Reviews: Use actions (like Codium or custom scripts) to auto-review PRs.
- Commit Messages: Use
git diff | llm "generate conventional commit"for standardized history. - Doc Generation: Auto-generate API docs from code signatures on merge.
