This directory is the deeper companion to AGENTS.md.
Use AGENTS.md for the one-screen AI-agent overview and this wiki for details that
are too specific or too changeable to keep at the repo root.
quickstart.md- canonical local validation commands and CI mapping.repo-map.md- where to edit and how the main subtrees relate.generated-files.md- derived outputs and their sources of truth.representations-and-bridges.md- the main polynomial representations and Mathlib bridge layers.typeclass-minimization.md- declaration-level typeclass discipline and minimal-assumption examples.binary-fields-and-ntt.md- the binary-field stack, GHASH model, and additive-NTT architecture.
AGENTS.mdis the canonical root guide for AI agents.CLAUDE.mdis only a symlink.- Keep one primary owner topic per page. The current pages are:
quickstart.mdfor commands, validation, and CI expectations.repo-map.mdfor repo structure and work-area routing.generated-files.mdfor derived outputs and source-of-truth rules.representations-and-bridges.mdfor representation choice and Mathlib bridges.typeclass-minimization.mdfor minimal typeclass assumptions and avoiding blanket instance scopes.binary-fields-and-ntt.mdfor the specialized field and NTT stack.
- Add new pages when a recurring topic no longer fits cleanly in an existing page.
- If a PR changes commands, repo structure, generated-file behavior, or recurring architecture guidance, update the matching page in the same PR.
- Keep these files committed so worktrees and delegated agents see the same guidance.
- Prefer linking to canonical docs over copying them.
../../README.md- public repo overview and primary exported surfaces.../../CONTRIBUTING.md- contribution process, naming, style, docstrings, and citations.../../ROADMAP.md- roadmap and current priorities.../../tests/README.md- test layout and commands.../../scripts/README.md- script inventory and operator notes.