Installable Agent Skills that load Arkiv know-how into your AI coding agent — Claude Code, Cursor, Codex. Each one is distilled from 124 real hackathon submissions and verified against the current SDK, so your agent ships the pattern instead of rediscovering the pitfall.
Start every build with data-modeling → expiration-design → query-patterns → write-safety. Then add your lane: privacy (anything user-sensitive), crosschain (contracts on another chain), agent-state (AI agents), realtime (live UX), and app-architecture before you deploy.
Design the entity schema before code: types, attributes vs payload, relationships, namespaces — from 124 real builds.
Design entity lifetime as schema — self-revoking grants, heartbeats, countdowns, and expiry as a feature.
Query Arkiv like the database it is — index at write time, paginate right, sort in JS, zero footguns.
Create, update, delete, extend, and transfer Arkiv entities without lost data, duplicate writes, or nonce collisions.
Entities are public — encrypt payloads client-side, index only safe attributes, never put secrets on-chain.
Arkiv as the queryable data layer next to your value chain — partition rule, mirror entities, chain indexers.
Wallet-owned, expiring agent state on Arkiv: tiered memory, recall queries, audit trails, leases.
Real-time UX on Arkiv is polling done well — subscriptions, cursor polls, SSE relays, presence via expiration.
Pick your Arkiv topology and signer model first, then ship: Next.js routes, env hygiene, deploy + prod checklists.
A skill is a folder with a SKILL.mdyour coding agent reads before writing Arkiv code. Copy a card's install prompt into your agent, or fetch the raw file yourself — every skill also ships a machine-readable skill-lock.json that powers the update radar above.
Ideas, not written yet. Each opens the case for building it — the gap it fills against the SDK, how it complements the live skills, and why it’s grounded.
Teaches your AI to stand up a new Arkiv project the right way: start from the official template, wire Braga, keep the signing key safe, and write a first create + query — using the MCP and SDK.
Teaches an AI the decision behind adding Arkiv to a web2 app: which records belong on Arkiv (public, verifiable, queryable), which stay in Postgres (private, hot, joined), and how to model the ones you move.
Teaches an AI how to test Arkiv apps — what to smoke-test on the testnet, how to structure fixtures, and how to assert TTL/expiry and full-replace behavior.
Teaches an AI to write Arkiv data cheaply — batch with mutateEntities, size payloads and attributes, tune TTL, update vs recreate.