./AGENTS.md

Sennin reads a repository‑level AGENTS.md to align reviews with your team’s context and constraints. Keep it short and actionable.
This file can also be used by other AI coding assistants like Codex, Cursor, GitHub Copilot, and others to stay aligned with your norms.

Cover these four sections

• Project overview — what this repo is for and success criteria.
  • 1–3 sentences: product, primary users, key outcomes.
  • Scope boundaries: what lives here vs in other services/packages.
  • Hard constraints (compliance, latency, availability, cost caps).
• Stack, code organization, architecture — high‑level map so Sennin lands in the right places.
  • Primary languages/frameworks and notable libs.
  • Directory layout and where domain logic lives.
  • State/data flow patterns; APIs used; error handling and logging approach.
  • Conventions worth following, and intentional deviations.
• Development principles — how to trade speed, safety, and design.
  • Bias: delivery speed vs long‑term quality; when each wins.
  • Testing policy (when to add tests, acceptable coverage, e2e vs unit).
  • Refactor policy (boy‑scout rules, when to defer cleanup).
  • Dependency policy (pinning, introducing new libs, browser/Node targets).
• Top unusual things & reviewer considerations — the gotchas.
  • Non‑standard practices.
  • Security/privacy hotspots to double‑check.
  • Concurrency/reliability requirements.
  • Release/ops constraints.