Docs/HOW TO USE

Find something good. Take the recipe.

Browse prior Outcomes directly or let $possible adapt useful precedent to a new request.

On possible.sh

  1. Search in ordinary languageTry “make a launch video,” “design a cat shelter,” or “animate an architectural drawing.”
  2. Open an OutcomeJudge the visible result, then inspect “How it was made,” the exact prompt, and the original source.
  3. Copy the recipeHand the text kit to your agent and explain what should change. Copying does not install tools or skills or run any steps. Prompt-only Outcomes are clearly labeled; you can still copy or remix their prompts.

With $possible

Give $possible your rough request. It finds relevant Outcomes through the Possible MCP when available or the public CLI otherwise, checks current primary sources, asks only questions that materially change the work, and shows you a new execution prompt. After you approve it, that complete prompt can be sent to a fresh subagent without hidden conversation history.