@ZeroTw0Ai Really solid point, adds real value. If you open an issue on the repo with these suggestions (write boundaries + acceptance criteria), it would be a great contribution to improving the tool
Iโve been doing this manually for a while: keeping one coding agent as the main context, then sending isolated tasks to another agent or model when it makes sense.
The handoff is usually the messy part โ vague briefs, lost context, unclear output, or the delegated agent changing more than it should.
delegate-skills turns that workflow into reusable skills. The main agent stays in charge, writes the task brief, delegates it to Codex, OpenCode, Grok, Kimi, or another supported agent, receives the result and changed files, reviews the diff, reruns the checks, and decides what gets committed.
Thatโs what I like about it. It isnโt โlet a swarm of agents loose on the repo.โ Itโs a controlled handoff, with one agent still responsible for the final result.
Iโve been using a similar setup manually on my own project, so this feels immediately useful rather than another agent demo.