epicenter

Shape a complex idea, workflow, or design around the judgment a person needs to make. Use when an explanation obscures its point or makes the user carry unnecessary complexity. Do not use for routine execution of a clear request.

Independently challenge a design and its underlying model to find deletion prizes and stronger invariants. Use when asking for an adversarial review, an independent architecture review, or an adversarial checkpoint on a plan or cumulative implementation. Do not use for routine final code inspection, which belongs to post-implementation-review.

Draft a `/goal` line that gives long-running Codex or Claude Code work an objective, validation evidence, and a stop condition. Use when the user explicitly asks for `/goal`, an agent goal, or a completion condition.

Write and maintain repository guidance across AGENTS.md, CLAUDE.md, and .agents/skills. Use when creating or editing skills, deciding where instructions belong, or diagnosing instruction sprawl and skill selection. Do not use merely because another skill is being executed.

Arktype: runtime validation, discriminated unions with .merge()/.or(), spread keys. Use when mentioning arktype, type(), union types, command/event schemas.

Identify an asymmetric win by refusing a small promise that would eliminate disproportionate complexity. Use when the user asks about asymmetric wins, pixel perfect fidelity, compatibility, fallbacks, modes, duplicate paths, or what would delete disproportionate complexity.

Epicenter auth packages: `@epicenter/auth` and the Svelte adapter at `@epicenter/auth/svelte`, OAuth sessions, identity state, auth-owned fetch/WebSocket, and the reload gate that makes a page lifetime one auth generation. Use when editing Epicenter auth clients, session state, hosted sign-in, or how a route boots from auth.

Autumn billing in Epicenter: `autumn.config.ts`, `autumn-js` credit checks, `atmn` CLI, plan gates, and metered AI usage. Use when changing billing, pricing, credits, plan access, refunds, or usage events.

Better Auth server/client setup: `auth.ts`, generated schema, DB adapters, sessions, cookies, env vars, and plugins. Use when mentioning Better Auth, betterauth, auth handlers, OAuth, email/password, or session configuration.

Better Auth security hardening: rate limits, secrets, CSRF, trusted origins, cookies, sessions, OAuth tokens, and audit logging. Use when reviewing auth security, brute-force protection, token handling, or deployment safety.

Show a proposed code change with before/after diffs, diagrams, or comparison tables. Use when the user asks to compare approaches or wants a visual proposal before a multi-file edit.

Cloudflare Workers patterns for Worker runtime APIs, Durable Objects, KV, R2, D1, Queues, WebSockets, streaming responses, bindings, wrangler configuration, and deployment limits. Use when users mention Cloudflare Workers, Durable Objects, KV, R2, D1, Queues, wrangler, or edge runtime behavior.

Find recurring Epicenter code smells and scope the cleanup they require. Use for periodic audits, cleanup PRs, post-refactor reviews, or reviews of a primitive’s consumers.

Apply Orwell's rules to the prose inside code — identifiers and comments — without changing behavior. Use when names, word choice, or comments need an editorial pass.

Audit Codex tasks and preserve unowned outcomes in `BACKLOG.md` while closing completed or superseded chats. Use when Codex task ownership or archive state needs cleanup, not generic issue-tracker or project-backlog work.

Collapse test-shaped production boundaries while preserving behavior and coverage. Use when a helper, wrapper, injected dependency, or export exists mainly to let a unit test reach internals.

Remove indirection that does not earn its boundary across a diff or package. Use when the user asks to run a collapse pass, simplify this broadly, audit dead abstractions, or shrink a surface, not when one function’s branching is the target.

Get Claude Code's second opinion on a proposal or implementation. Use when the user asks to consult Claude, requests Claude's design judgment or research, or invokes a design-review workflow that includes Claude. Do not enlist Claude merely because a task is complex.

Linearize complex control flow with early returns, guard clauses, nested ifs, and natural-language conditions. Use when flattening nested conditions, linearizing try/catch logic, or separating mixed throw-and-return handlers, not for a broad "simplify this" pass over a diff or package, which belongs to collapse-pass.

defineErrors from wellcrafted: variant factories, extractErrorMessage, InferErrors/InferError, call site patterns. Use when creating error types or reviewing error patterns.