freeform-data

A set of low-level helpers for asking AI models to produce text, code, Markdown, or loose JSON in an Elixir application. It centralises provider settings and keeps provider requests separate from task-specific code.

In plain words
What is it for?
Use it when building Elixir features that need free-form AI responses rather than data validated against a strict schema.
Why use it?
It gives different AI features one place for credentials and model names while allowing each feature to shape its own unvalidated output.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/fullstack-phoenix/saas_kit/freeform-data
Any agent
npx skills add fullstack-phoenix/saas_kit --skill freeform-data
Clone the repo
git clone --depth 1 https://github.com/fullstack-phoenix/saas_kit

Made for: Claude Code, Codex.

Per session 34 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,074 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00034 $0.02074
Opus 5 $0.00017 $0.01037
Sonnet 5 $0.00007 $0.00415
Haiku 4.5 $0.00003 $0.00207

Measured yesterday against content hash c4317aab2dfd, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

freeform-data scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/ai-integration/freeform-data/SKILL.md · 263 lines

How it starts

The opening of the file, as written. The whole thing — 263 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Freeform Data

Use this skill when an AI feature should return text, code, markdown, or loose JSON instead of a validated schema.

Goal

  • Reuse the shared MyApp.Agents configuration.
  • Keep a low-level MyApp.Agents.LLM module for provider-specific requests.
  • Let task-specific modules build on top of that helper.
  • Keep tests isolated with Mimic.

Root Agents Module

Use one shared root module for API keys and model names:

defmodule MyApp.Agents do
  @moduledoc """
  Central configuration for AI agents.
  """

  def anthropic_key, do: Application.get_env(:my_app, :anthropic)[:api_key]
  def openai_key, do: Application.get_env(:my_app, :openai)[:api_key]

  def model(:haiku), do: "claude-haiku-4-5"
  def model(:sonnet), do: "claude-sonnet-4-5"
  def model(:gpt_5_mini), do: "gpt-5-mini"
  def model(:gpt_5), do: "gpt-5.2"
  def model(:gpt_4_1_mini), do: "gpt-4.1-mini"
end

Credential Placement

Do not put provider credentials in shared config/config.exs. Keep local credentials in config/dev.exs:

config :my_app, :anthropic, api_key: System.get_env("ANTHROPIC_API_KEY")
config :my_app, :openai, api_key: System.get_env("OPENAI_API_KEY")

Low-level LLM Module

Keep one shared helper module for raw LLM calls:

defmodule MyApp.Agents.LLM do
  import MyApp.Agents

  @claude_url "https://api.anthropic.com/v1/messages"
  @openai_url "https://api.openai.com/v1/chat/completions"

  def claude(content, opts \\ [])

  def claude("" <> content, opts) do
    [%{role: "user", content: content}]
    |> claude(opts)
  end

  def claude(messages, opts) when is_list(messages) do
    model_name = Keyword.get(opts, :model, model(:haiku))
    max_tokens = Keyword.get(opts, :max_tokens, 4_096)
    extract_code = Keyword.get(opts, :extract_code, false)
    extract_json = Keyword.get(opts, :extract_json, false)

    Req.post(@claude_url,
      headers: [
        {"Content-Type", "application/json"},
        {"x-api-key", anthropic_key()},
        {"anthropic-version", "2023-06-01"}
      ],
      receive_timeout: 90_000,
      json: %{
        max_tokens: max_tokens,
        model: model_name,
        messages: messages
      }
    )
    |> parse()
    |> maybe_extract_code(extract_code)
    |> maybe_extract_json(extract_json)
  end

  def openai(content, opts \\ [])

  def openai("" <> content, opts) do
    [%{role: "user", content: content}]
    |> openai(opts)
  end

  def openai(messages, opts) when is_list(messages) do
    model_name = Keyword.get(opts, :model, model(:gpt_5_mini))
    reasoning_effort = Keyword.get(opts, :reasoning_effort, "medium")
    max_tokens = Keyword.get(opts, :max_tokens, 4_096)
    extract_code = Keyword.get(opts, :extract_code, false)
    extract_json = Keyword.get(opts, :extract_json, false)
    prediction = Keyword.get(opts, :prediction)

    Req.post(@openai_url,
      headers: [
        {"Content-Type", "application/json"},
        {"Authorization", "Bearer #{openai_key()}"}
      ],
      receive_timeout: 90_000,
      json:
        %{
          model: model_name,
          max_completion_tokens: max_tokens,
          messages: messages
        }
        |> maybe_add_reasoning_effort(reasoning_effort, model_name)
        |> maybe_add_prediction(prediction)
    )
    |> parse()
    |> maybe_extract_code(extract_code)
    |> maybe_extract_json(extract_json)
  end

  defp maybe_add_prediction(json, "" <> content) do
    json
    |> Map.delete(:max_completion_tokens)
    |> Map.put(:prediction, %{content: content, type: "content"})
    |> Map.put(:model, model(:gpt_4_1_mini))
  end

  defp maybe_add_prediction(json, _), do: json

  defp maybe_add_reasoning_effort(json, "" <> reasoning_effort, model_name)
       when model_name in [model(:gpt_5), "o4-mini", "o3", "o3-mini"] do
    Map.put(json, :reasoning_effort, reasoning_effort)
  end

  defp maybe_add_reasoning_effort(json, _, _), do: json

  defp parse({:ok, %{body: %{"content" => [content | _]}}}) do
    case content do
      %{"text" => text, "type" => type} -> {:ok, %{content: text, type: type}}
      _ -> {:error, %{content: "No result"}}
    end
  end

  defp parse({:ok, %{body: %{"choices" => [choice | _]}}}) do
    case choice do
      %{"message" => %{"content" => content}} -> {:ok, %{content: content}}
      _ -> {:error, %{content: "No result"}}
    end
  end

  defp parse(_), do: {:error, %{content: "No result"}}

  defp maybe_extract_code({:ok, %{content: content}}, true) do
    {:ok, %{content: extract_code(content)}}
  end

  defp maybe_extract_code(result, _), do: result

  defp maybe_extract_json({:ok, %{content: content}}, true) do
    {:ok, %{content: extract_json(content)}}
  end

  defp maybe_extract_json(result, _), do: result

  defp extract_code(content) do
    content
    |> String.replace("```elixir", "```")
    |> String.replace("```javascript", "```")
    |> String.replace("```json", "```")
    |> String.trim()
    |> String.split("```")
    |> case do
      [_, code | _] -> code
      [text] -> text
    end
  end

  def extract_json(text_or_json) do
    text_or_json
    |> String.replace("```json", "```")
    |> String.trim()
    |> String.split("```")
    |> case do
      [_, json | _] -> json
      [json] -> json
    end
    |> Jason.decode!()
  rescue
    _ -> %{}
  end
end

Read the full file on GitHub · 263 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 263 lines · 34 tokens per session scan A c4317aab2dfd

Subscribe to this mod's changes

freeform-data is a skill published in the GitHub repository fullstack-phoenix/saas_kit (2 stars, last pushed 2mo ago), licensed MIT. It adds 34 tokens to every session and 2,074 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openai/codex · 114 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openai/codex · 113 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens