aigorahub/elves — reverse-engineered prompt

Build me a developer tool called Elves that lets Claude Code or Codex work through a big coding plan while I’m away. I want it to feel like a skill I can drop into a repo, give it a plan, stage the run first, then start it in a second prompt so the agent can work unattended without losing track.

It should guide the agent through batches of work, run tests, check the repo is still green, review its own changes, respond to PR comments, update docs, push checkpoints, and keep going until the plan is done or it hits a real blocker. Please include the durable memory pieces so it can survive context resets, like a survival guide, learnings file, execution log, plan, and session state.

Also add a simple PR landing command that does one final review pass, handles feedback, waits for checks, and merge commits when everything is clean. At the end of a substantial run, generate a readable HTML report explaining what happened while I was away.