Adaptivetest Skill

Adaptive testing engine with IRT/CAT, AI question generation, and personalized learning recommendations

Security Scan

The skill's requirements and runtime instructions are consistent with an API wrapper for an adaptive testing service and do not request unrelated system access or secrets.

Purpose & Capability

Name/description map directly to the documented API endpoints and capabilities (IRT/CAT, question generation, recommendations). The only required artifact (ADAPTIVETEST_API_KEY) and required binary (curl) are appropriate for an HTTP API client.

Instruction Scope

SKILL.md contains only HTTP endpoint usage and API workflows that align with the stated purpose. It does not instruct reading local files, shell history, or unrelated environment variables, nor does it direct data to endpoints other than the documented base URL.

Install Mechanism

No install spec or downloaded code is present; this is an instruction-only skill that assumes an HTTP client (curl) is available. That minimizes code written to disk and reduces install-time risk.

Credentials

The skill requires a single API key (ADAPTIVETEST_API_KEY) to authenticate to the documented service. That is proportionate to an API-wrapping skill; no unrelated credentials or config paths are requested.

Persistence & Privilege

The skill is not configured as always-on and is user-invocable. It does not claim any ability to modify other skills or system-wide agent settings.

Assessment

This skill appears coherent with its described purpose, but before installing consider: (1) only provide an API key from a trusted AdaptiveTest account and avoid using high-privilege or production keys when testing; (2) confirm the vendor, domain (the skill points to a Railway-hosted production URL), and data-handling promises (FERPA/PII protections, data residency, retention, and deletion policies) meet your legal/compliance needs; (3) test with non-sensitive or sandbox data first and monitor rate limits; (4) rotate keys regularly and scope them (least privilege) if the platform supports it; and (5) review the provider's SLA, pricing, and support/contact info (README/CLAUDE.md include a contact email). If any of these details are unclear from the vendor, ask for clarifying documentation before sending real student data.

AdaptiveTest

Production-grade adaptive testing API. Uses Item Response Theory (IRT 2PL/3PL) with Computerized Adaptive Testing (CAT) to deliver precise ability estimates in fewer questions. Includes AI-powered question generation and personalized learning recommendations.

When to Use This Skill

Use AdaptiveTest when the user needs to:

• Create or manage assessments and tests
• Run adaptive testing sessions that select questions based on student ability
• Generate assessment questions by topic, difficulty, or academic standard
• Get personalized learning recommendations for students
• Calibrate test items using IRT parameter estimation
• Manage students, classes, and enrollments
• Analyze test results and track student mastery

Authentication

All requests require the X-API-Key header:

X-API-Key: ${ADAPTIVETEST_API_KEY}

Base URL: https://adaptivetest-platform-production.up.railway.app/api

Core Workflows

1. Create and Administer an Adaptive Test

POST /tests              -- Create a test (set cat_enabled: true)
POST /tests/{id}/items   -- Add items to the test
POST /tests/{id}/sessions -- Start an adaptive session for a student
GET  /sessions/{id}/next-item -- Get the next CAT-selected item
POST /sessions/{id}/responses -- Submit student response
GET  /sessions/{id}/results   -- Get ability estimate and results

The CAT engine selects items using maximum Fisher information. Ability is estimated after each response using IRT 2PL or 3PL models. Sessions terminate when the standard error drops below threshold or max items are reached.

2. Generate Questions with AI

POST /gen-q -- Generate questions by topic, difficulty, and standard

Request body:

{
  "topic": "Quadratic equations",
  "difficulty": "medium",
  "count": 5,
  "standard": "CCSS.MATH.CONTENT.HSA.REI.B.4",
  "format": "multiple_choice"
}

Returns QTI 3.0-compatible items with stems, distractors, and rationales. Generation takes ~7 seconds.

3. Get Learning Recommendations

POST /recs -- Get personalized learning recommendations for a student

Request body:

{
  "student_id": "student-uuid",
  "subject": "Mathematics",
  "include_resources": true
}

Returns a personalized learning plan based on the student's ability profile and assessment history. Generation takes ~25 seconds.

4. Calibrate Test Items

POST /tests/{id}/calibrate -- Run IRT calibration on collected response data

Requires sufficient response data (minimum 30 responses per item recommended). Returns IRT parameters: difficulty (b), discrimination (a), and guessing (c) for 3PL.

5. Manage Students and Classes

POST /students           -- Create a student
GET  /students           -- List students
POST /classes            -- Create a class
POST /classes/{id}/enroll -- Enroll students in a class

OneRoster 1.2 compatible for SIS integration.

6. View Results and Analytics

GET /sessions/{id}/results       -- Detailed session results with ability estimate
GET /students/{id}/history       -- Assessment history for a student
GET /tests/{id}/analytics        -- Item-level analytics for a test

Rate Limits

Rate limits depend on your API key tier. Check X-RateLimit-Remaining header on each response.

Error Handling

All errors return JSON with a detail field:

{"detail": "Human-readable error message"}

Common status codes: 400 (validation), 401 (auth), 403 (limit exceeded), 404 (not found), 429 (rate limited).

Reference Documentation

For detailed endpoint specifications, request/response shapes, and IRT/CAT concepts, see the references/ directory:

• references/api-endpoints.md -- Full endpoint reference
• references/adaptive-testing.md -- IRT and CAT concepts
• references/calibration.md -- Item calibration guide