Leonxlnx/unlazy ? reverse-engineered prompt

Build me a JavaScript skill for AI coding agents called unlazy that helps them do substantial work without cutting corners.

I want it to make the agent write an acceptance ledger first, run only the approved checks, reverify the work before claiming it is done, and keep the output honest about what the evidence actually shows. It should support a depth based workflow where a task can be split into layers, with clear gate files, claims, ownership, and simple status tracking for leaves and branches. Include the checker that reads a gate file, approves commands, runs them safely, and records evidence. Also include the optional Claude Code hook that can block the session until the ledger is satisfied.

Please make the install and usage feel simple, with clear docs, templates, examples, and sensible defaults. If anything needs current agent or CLI details, look up the latest docs online first.