Your agent follows the skills
clue init installs six skills and an AGENTS.md routing hub. Any coding agent that reads them knows how to plan, make a change, verify it, and upgrade, so your prompts can stay in ordinary words.
Coding agents can produce plausible changes faster than a person can confidently accept them. Cliewen keeps the requirement, implementation, and acceptance evidence connected in Git, then checks that the connection is intact before merge.
Pick up any artifact and follow it back to why it exists, or forward to what proves it. clue checks that no arrow is missing. It does not run your tests, it cannot tell you that a test proves the right thing, and it cannot tell you the goal was worth having. That judgment stays yours, at the merge.
.agents/skills/: the process your coding agent follows.docs/: the lasting record of the system as it exists.clue: the command-line judge that checks the corpus and its evidence.Install takes one command. Prompting takes ordinary words. The only step Cliewen refuses to do for you is the last one.