agentsclimarketplace

Ash framework

Skill oliver-kriska/claude-elixir-phoenix/targets/codex/skills/ash-framework

Claude Code plugin for Elixir/Phoenix/LiveView — 20 specialist agents, Iron Laws enforcement, and Tidewave MCP integration. Plan features with parallel research agents, execute with automatic verification, review with 4-agent parallel audits, and capture learnings as reusable knowledge.

Install
npx -y skills add oliver-kriska/claude-elixir-phoenix --skill ash-framework

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Ash Framework — resources, actions, policies, aggregates; Use when generating resources via mix ash.codegen, editing…

SKILL.md

5.2 KB, as published. Nobody here has run it

Ash Framework Reference

Reference for Ash Framework in Phoenix/LiveView projects. Ash complements Phoenix/Ecto — LiveView, security, and OTP Iron Laws still apply. Only data access patterns shift toward Ash actions and domain code interfaces.

Iron Laws

  1. USE DOMAIN CODE INTERFACES — Never call Ash.create/Ash.read directly in LiveViews or Controllers; use domain code interfaces: MyApp.Accounts.register_user() not Ash.create(User, attrs)
  2. SET ACTOR/SCOPE AT QUERY PREP, NOT EXECUTION — Pass actor: or scope: to for_read/for_create/for_action (prep), NOT to Ash.read!/Ash.create! (execution); execution-level actor bypasses row-level policy evaluation. If project uses Ash.Scope, pass scope: consistently instead of bare actor: — do not mix styles
  3. GENERATORS FIRST — Before writing Ash code manually, run mix ash.gen.resource or mix ash.gen.domain with --yes; check mix help ash.gen.<task> for options
  4. CODEGEN AFTER RESOURCE CHANGES — Always run mix ash.codegen after modifying resources; this generates migrations from resource snapshots — never write AshPostgres migrations by hand
  5. ACTIONS OVER FUNCTIONS — Put business logic in named actions, not domain functions; expose via code interfaces defined on the domain
  6. NEVER EDIT RESOURCE SNAPSHOTSpriv/resource_snapshots/ is owned exclusively by mix ash.codegen; manual edits corrupt migration tracking
  7. NO DIRECT Repo.* IN ASH PROJECTSRepo.all/get/insert bypass Ash policies and notifications; use domain code interfaces. Any Repo call in an Ash project is an escape hatch and must be documented

Quick Reference

Domain Code Interface Pattern

# Domain definition
defmodule MyApp.Accounts do
  use Ash.Domain

  resources do
    resource MyApp.Accounts.User do
      define :register_user, action: :create, args: [:email, :password]
      define :get_user_by_email, action: :read, get_by: [:email]
    end
  end
end

# In LiveView/Controller — always via domain, never Ash.create directly
{:ok, user} = MyApp.Accounts.register_user(email, password, actor: nil)
user = MyApp.Accounts.get_user_by_email!(email, actor: current_user)

Authorization — Actor/Scope at Query Prep

# CORRECT — actor at query prep, policies evaluated per-row
MyApp.Post
|> Ash.Query.for_read(:list_published, %{}, actor: current_user)
|> Ash.read!()

# CORRECT with Ash.Scope (carries actor + tenant + context; use if project adopts it)
MyApp.Post
|> Ash.Query.for_read(:list_published, %{}, scope: scope)
|> Ash.read!()

# WRONG — actor at execution bypasses row-level policy evaluation
MyApp.Post
|> Ash.Query.for_read(:list_published)
|> Ash.read!(actor: current_user)

Ash.Scope — When the Project Uses It

Ash.Scope bundles actor + tenant + context into a single struct passed through actions. Implement Ash.Scope.ToOpts on a project-defined scope struct:

defimpl Ash.Scope.ToOpts, for: MyApp.Scope do
  def get_actor(%{current_user: u}), do: {:ok, u}
  def get_tenant(%{current_tenant: t}), do: {:ok, t}
  def get_context(%{locale: l}), do: {:ok, %{shared: %{locale: l}}}
  def get_tracer(_), do: :error
  def get_authorize?(_), do: :error
end

Detection: if the project has a Scope module implementing Ash.Scope.ToOpts, use scope: everywhere instead of bare actor:. Do NOT mix the two styles in the same codebase. See mix usage_rules.docs Ash.Scope for full protocol spec.

File Conventions (from mix ash.gen.*)

FileLocationBehaviour
Changeslib/app/ctx/changes/name.exuse Ash.Resource.Change
Policy Checkslib/app/ctx/checks/name.exuse Ash.Policy.Check
Custom Actionslib/app/ctx/actions/name.exgeneric action logic
Custom Typeslib/app/ctx/types/name.exuse Ash.Type
Validationslib/app/ctx/validations/name.exuse Ash.Resource.Validation

Generator Workflow

mix ash.gen.resource MyApp.Accounts.User --yes
mix ash.gen.domain MyApp.Accounts --yes
mix ash.codegen        # reads resource snapshots → generates migration
mix ash.migrate

Research

Prefer the highest-fidelity source available:

  1. Tidewave (exact version from mix.lock):

    mcp__tidewave__get_docs(module: "Ash.Resource")
    mcp__tidewave__get_docs(module: "AshPhoenix.Form")
    
  2. usage_rules (project-synced to your installed ash_* dep versions):

    mix usage_rules.search_docs "<topic>" -p ash -p ash_phoenix -p ash_postgres -p ash_authentication -p ash_oban
    mix usage_rules.docs Ash.Resource
    
  3. WebFetch hexdocs.pm (fallback when neither is available):

    WebFetch(url: "https://hexdocs.pm/ash/Ash.Resource.html", prompt: "Extract module docs.")
    

If usage_rules is not configured, the SessionStart hook suggests how to install it.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.