How to Write an AGENTS.md File for AI Coding Agents
Build a concise repository contract for AI coding agents with scoped instructions, exact commands, architecture boundaries, review rules, and a practical Turborepo example.
Read the field guideDetailed field guides for debugging Next.js, Sanity, deployment, caching, and the systems around them.
Read the latest guide
NEWAI agentsBuild a concise repository contract for AI coding agents with scoped instructions, exact commands, architecture boundaries, review rules, and a practical Turborepo example.
Read the field guide
02Ad Tech · Next.jsCapture GCLID, GBRAID, and WBRAID safely, preserve attribution through a CRM, and upload qualified lead conversions with Google Data Manager API.
Read the field guide
03Next.js · Ad TechBuild reliable Next.js conversion tracking for Meta CAPI and Google Ads with consent, deduplication, durable delivery, platform-specific hashing, and reconciliation.
Read the field guide
04UID2 · Ad TechBuild a reliable UID2 audience pipeline: normalize email and phone data correctly, map identities, honor opt-outs, and activate through participating DSPs.
Read the field guide
05Next.js · DebuggingConfigure Next.js Proxy matchers that exclude static assets and prefetch requests, protect the intended routes, and stay testable in production.
Read the field guide
06Doppler · VercelUse Doppler as the source of truth for secrets, map configs to Vercel environments, and rotate credentials without configuration drift.
Read the field guide
07Sanity · Next.jsDiagnose and fix Sanity API request spikes, repeated RSC prefetches, and Vercel ISR writes after moving a Sanity Live app to Next.js 16.
Read the field guide
08HubSpot · SecurityReplace Base64 email parameters with stable internal IDs, short-lived signed click tokens, clean redirects, limited cookies, and explicit trust boundaries.
Read the field guideDev Fieldnotes publishes focused guides built around real failure modes, tested configurations, and verification steps you can use in production.
Focused starter kits built from real implementation work—tested, documented, and designed to skip the fragile first draft.
Explore the kitsDevelopment-only metadata discovery
Entities, operations, fields, and limits
Validated output with runtime guardrails