Claude Code Skills and Subagents: How I Use Both
Skills are reusable instructions that load into your context; subagents run in their own. Here's when a solo dev reaches for each, with real examples.
Skills are reusable instructions that load into your context; subagents run in their own. Here's when a solo dev reaches for each, with real examples.
Structure a project so AI coding tools write correct code: a root AGENTS.md, shallow folders by responsibility, task files, and machine-checkable guardrails.
A practical checklist for reviewing AI-generated code when you are the only reviewer: scope, correctness, security, migrations, dependencies, reuse, and tests.
A deep dive into automating Google Ads with Claude Code and the Google Ads API: campaign hierarchy, micros budgets, GAQL reporting, and a PAUSED safety gate.
Automating Meta (Facebook/Instagram) ads with Claude Code and the Marketing API: cents budgets, targeting, image_hash creatives, and a PAUSED safety gate.
A two-layer Claude Code setup for marketing automation: skills handle judgment, Python scripts make every API call, and humans approve before any spend.
I built an AI writing workflow in Claude Code and kept reaching for software engineering: single source of truth, single-responsibility skills, and test-first goals.
A practical Claude Code workflow for solo developers: scope the task, write a task file, ask for a plan before code, review the diff, and let the agent review its own work.