Set up a repository whose rules are checked

Set up a new or existing repository so that its rules are checked at every edit, commit and push. The repository says its jobs as capability rows, from what each folder's README says it does. Its DATA files are pinned by content. Its retired paths are declared. It is registered with the person's own checks, and those checks are shown to pass on it. Every input is the repository's own files or the person's own check settings. The AI writes no job the repository's own README does not say, and changes no check to make it pass.

What it asks for

  • pin_file (Optional, words): The pin file for that folder, under the repository
  • registries (Needed, list): The person's own check registry files, each with its own schema: where the repository is declared, which commit checks run on it, which folders an agent may write in
  • repository (Needed, words): The repository folder
  • unpinned_ok (Optional, list): Files in the product folder that are never pinned
  • check_command (Needed, words): The person's own command that runs their commit checks on a repository's staged files
  • hook_installer (Optional, words): The person's own command that installs their git hooks in a repository, if they have one
  • product_folder (Optional, words): The folder of DATA files a program reads at run time, under the repository, when it has one

The steps your AI walks

  1. Read what the repository says it is
  2. Write the repository's capability rows and places
  3. Pin the DATA files by content
  4. Declare the paths the repository retired
  5. Register the repository with the person's checks
  6. Show the checks pass on the repository

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this workflow by what it does and runs it.

Say to your AI: “Use Trillion to set up a new or existing repository so that its rules are checked at every edit, commit and push. The repository says its jobs as capability rows, from what each folder's README says it does. Its DATA files are pinned by content. Its retired paths are declared. It is registered with the person's own checks, and those checks are shown to pass on it. Every input is the repository's own files or the person's own check settings. The AI writes no job the repository's own README does not say, and changes no check to make it pass”