PlugWiki

TypeSafe AI Skill for Claude: Build with Typed AI Decisions

Guides Claude when you write code that calls TypeSafe's Jev model for typed answers with probabilities.

Installs
98.5K
on skills.sh
Stars
2.9K
on GitHub
Files
1
besides SKILL.md
Updated
Oct 2026
last repo commit

typesafe-ai teaches Claude how to build with TypeSafe, an API whose Jev model answers typed questions about text, images and app data. Answers come back as typed data with probabilities, for example a Choice among options, the probability that a statement is true, a Score, or a box on an image. The skill steers Claude to let Jev make the judgments and keep the final decision in your code.

It loads when you build, review or debug code that calls TypeSafe, or when you want to replace a prompt-and-parse step or a keyword heuristic with a typed question. Claude reads TypeSafe's live docs for syntax and limits. Test requests to the live API use your TYPESAFE_API_KEY, and the skill tells Claude to ask your permission first.

Good for

  • Developers calling the TypeSafe API
  • Teams replacing an LLM prompt-and-parse step with typed answers
  • Anyone exploring where typed yes/no judgments fit an app

Install typesafe-ai

Claude Code

  1. 1

    Copy this prompt.

    Install the typesafe-ai skill in Claude Code. Run this command:

claude plugin marketplace add typesafe-ai/skills
claude plugin install typesafe@typesafe-ai

Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

The README gives the plugin route for Claude Code and says you can then invoke the skill as /typesafe:typesafe-ai. The skills CLI asks which agent to install for and installs per project unless you add -g. Pick one route, not both.

From the project README. Source.

What's in the skill

When it loads

TypeSafe integration craft for System One models (sometimes called Decision Models) such as Jev, which answer typed questions about text, images, and application state with calibrated probabilities. Use when building, reviewing, or debugging code that calls TypeSafe; when writing its questions or state; when replacing an LLM prompt-and-parse step, keyword matching, or a brittle heuristic with a typed question; or when brainstorming where programmable common sense fits in an app.

Sections in SKILL.md

  1. Judgments from Jev, decisions in code
  2. Live docs
  3. Workflow
  4. Design the state
  5. Write the questions
  6. Criteria
  7. Compose requests
  8. Read the answers
  9. Confidence is rarely the gate
  10. Probabilities and thresholds
  11. Test with the live API
  12. When an answer looks wrong
  13. Yellow flags
  14. Cookbooks

Key facts

Author
TypeSafe (typesafe-ai)
Repository
typesafe-ai/skills
Installs on skills.sh
98,482 (crawled 2026-11-02)
GitHub stars
2,863 (whole repo)
License
MIT
Last repo commit
2026-10-23
Files in the skill folder
2 files: SKILL.md and LICENSE. No scripts.
skills.sh audits
Gen Agent Trust Hub: pass, Socket: pass, Snyk: pass
Works in (per README)
Claude Code (plugin), and other agents through the skills CLI

What it includes

  • Sketch the rules or truth table first, then list the judgments they need

    Start from the decision logic

  • Pick what text, images and data each question sees; check it with a human test

    Design the state

  • One concrete judgment per question; split questions joined by "or"

    Write the questions

  • Bundle questions that share the same state into one request

    Compose requests

  • Use probabilities and thresholds, not confidence; 0.7 and 0.3 for yes/no to start

    Read the answers

  • Sends sample requests with your TYPESAFE_API_KEY after you approve budget and data

    Network · Test with the live API

  • Reads docs.typesafe.ai pages for request syntax, SDKs, question types and limits

    Network · Look up the live docs

  • Five starting patterns, from routing and filling arguments to verify-and-escalate checks

    Cookbooks

From the official docs

Permissions

Low risk

Test requests send your text, images or app data to the TypeSafe API using your key and your experiment budget. Approve each test run only after you have seen the payload and the budget Claude proposes.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • It asks before calling the live API

    SKILL.md tells Claude to get your permission, agree a budget and disclose what data will be sent, pointing out any personal or private data. Say no if the data should not leave your machine.

  • Test requests use your TypeSafe key

    Claude needs network access to the API endpoint and your TYPESAFE_API_KEY to run them.

  • The plugin route adds a third-party marketplace

    The README installs it with claude plugin marketplace add typesafe-ai/skills, then claude plugin install typesafe@typesafe-ai. The plugin source is the repo root. The skills CLI route is the alternative.

Technical details
Talks todocs.typesafe.ai (live docs). api.typesafe.ai (test requests)
Sign-in and keysSKILL.md tells Claude to say that your TYPESAFE_API_KEY will be used for test requests and to keep API keys on the server. It does not describe how the key is stored.
Scopes

  • TYPESAFE_API_KEY (live test requests)

Questions

What does the typesafe-ai skill do?

It makes Claude follow TypeSafe's guidance when you write code that calls TypeSafe. That covers designing the state each question sees, writing one-judgment questions, reading probabilities and testing against the live API. Decisions stay in your code.

How do I install it?

Use the install box at the top of this page. The README also documents a Claude Code plugin route and says you can invoke the skill by name with /typesafe:typesafe-ai.

Do I need a TypeSafe API key?

Only for live test requests. The skill tells Claude to ask your permission first, agree a budget, tell you what data will be sent and say that your TYPESAFE_API_KEY will be used.

Is it safe to use?

The skill folder holds only SKILL.md and a LICENSE, so it bundles no scripts. The skill tells Claude to ask before sending test data to the API, to point out any personal or private data, and to keep API keys on the server. Review the payloads before you approve a test run.

Does TypeSafe make the final decision in my app?

No. The skill says integrations tend to work best when Jev makes the judgments and your code decides what to do with them, like a truth table written in code.

What should I ask Claude to try first?

The README suggests: "Use TypeSafe to route incoming support tickets by department, with human review for uncertain decisions."

Information

updated Nov 2, 2026

Repository
typesafe-ai/skills
Folder
skills/typesafe-ai

typesafe-ai/skillsskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2
  1. typesafe-ai/skills README · github.com
  2. typesafe-ai SKILL.md · github.com

Updated Nov 2, 2026.