Review of delegate-skills
Most AI coding tools force you to context-switch. You stop what you're doing, switch to the AI, watch it work, then switch back. delegate-skills takes a different approach: it lets the AI work in the background while you stay in flow.
You write a brief spec for what you want implemented (markdown, free-form). delegate-skills hands it to the OpenAI Codex CLI in a separate worktree. Codex implements while you keep coding in your main branch. When you have a break, you switch to the worktree, review the diff, accept or request changes. If accepted, the changes merge into your main branch.
The cognitive cost of context-switching is real. Most AI tools don't account for it. delegate-skills treats the AI as a junior developer you can hand off work to, rather than a tool you operate. The result is more work done per day, less mental fatigue.
Self-contained features: adding a new API endpoint, writing a test file, generating boilerplate. Tasks with clear specs. Tasks that don't need deep context from your current code (Codex can read the repo, but it doesn't have your in-flight context).
A useful pattern for solo developers and small teams. The skill itself is small - the value is in the workflow it enables. If you find yourself stopping work to do an AI task, try delegating instead.
|