Delegate planned agent work with Elves
Use Elves to hand a bounded development or research plan to a separate Claude Code or Codex worker while preserving run files, worktree safety, review gates, and human merge authority.
npx skills add agentskillexchange/skills --skill delegate-planned-agent-work-with-elves
Elves is an open-source Agent Skill for delegating planned development or research work to a separate worker session. The main Claude Code or Codex driver stages the run, creates durable run files, prepares a branch or dedicated worktree when needed, launches a native worker, reviews cumulative progress, and stops at a landable PR or reviewed result while the user keeps merge authority.
What this skill actually does
Use this when a human has a concrete plan and wants an agent to carry the middle of the run through implementation, proof, and review without tying the main session to a single provider. The boundary is bounded delegation with explicit plan acceptance, run memory, safety checks, and final review. It is not a generic agent framework, desktop app, or permissionless auto-merge bot.
Inputs and prerequisites: Claude Code or Codex, Python 3, git, optional GitHub CLI or configured worker adapters.
Setup notes: Clone https://github.com/aigorahub/elves, then run scripts/sync_installed_skills.py –apply –target claude for Claude Code or –apply –target codex for Codex. Validate with the installed install_doctor.py startup check.
Source and verification boundary: use https://aigorahub.github.io/elves/ as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.