AI Text Detector Agent Skill — AI Pass API
Overview
AI Pass "AI Text Detector" (app slug: ai-detector) is a detection skill that analyzes input text and classifies whether it was written by an AI or by a human. The skill returns a classification label (e.g., "AI-generated" or "Human-written"), a confidence score (0.0–1.0), and a short explanation of the signals that influenced the decision.
Use cases for AI agents:
- Content moderation: flaging or routing likely AI-generated content for review.
- Authorship verification: assisting editors or educators to detect AI-written submissions.
- Metadata enrichment: adding a detection label and confidence to documents in a pipeline.
- Conversational agents: determining whether a message may be synthesized and adapting response style or verification steps.
Skill File (YAML)
name: ai-detector
description: Detects whether text was written by AI or a human
version: "1.0"
auth:
type: bearer
token_env: AIPASS_API_KEY
endpoints:
- name: generate
method: POST
url: https://aipass.one/apikey/v1/chat/completions
headers:
Authorization: "Bearer $AIPASS_API_KEY"
Content-Type: application/json
Example Usage
Curl example
curl -s https://aipass.one/apikey/v1/chat/completions \
-H "Authorization: Bearer $AIPASS_API_KEY" \
-H "Content-Type: application/json" \
-H "X-Client-ID: YOUR_CLIENT_ID" \
-d '{
"model": "aipass-detector-v1",
"messages": [
{"role": "system", "content": "You are an AI detector. For the provided user text, return JSON with keys: label (\"AI\" or \"Human\"), confidence (0.0-1.0), explanation (short)."},
{"role": "user", "content": "Paste the text to analyze here."}
],
"temperature": 0
}'
Example response (abridged)
{
"id": "resp-xyz",
"choices": [{
"message": {
"role": "assistant",
"content": "{\"label\":\"AI\",\"confidence\":0.92,\"explanation\":\"consistent phraseology, low lexical diversity, and high token probability patterns.\"}"
}
}]
}
Python example
import os
import requests
import json
api_key = os.getenv("AIPASS_API_KEY")
url = "https://aipass.one/apikey/v1/chat/completions"
headers = {
"Authorization": f"Bearer {api_key}",
"Content-Type": "application/json",
"X-Client-ID": "YOUR_CLIENT_ID"
}
payload = {
"model": "aipass-detector-v1",
"messages": [
{"role": "system", "content": "You are an AI detector. For the provided user text, return JSON with keys: label (\"AI\" or \"Human\"), confidence (0.0-1.0), explanation (short)."},
{"role": "user", "content": "The text to analyze goes here."}
],
"temperature": 0
}
resp = requests.post(url, headers=headers, json=payload)
resp.raise_for_status()
data = resp.json()
# The detector returns its result as a JSON string inside the assistant message content.
assistant_msg = data["choices"][0]["message"]["content"]
result = json.loads(assistant_msg)
print("Label:", result["label"])
print("Confidence:", result["confidence"])
print("Explanation:", result["explanation"])
Notes:
- Always include Authorization: Bearer $AIPASS_API_KEY. Never embed a real API key into code or public documentation.
- Use temperature: 0 or a low value for consistent deterministic classification.
- Include X-Client-ID: YOUR_CLIENT_ID where required by your integration.
Available Models
- aipass-detector-v1 — Primary detection model tuned for high precision on synthetic text patterns and robust confidence scoring.
- aipass-detector-lite — Lightweight, lower-latency model for bulk or realtime checks with slightly reduced accuracy.
- aipass-ensemble-v1 — Higher-accuracy ensemble detector that combines multiple signals (token patterns, style features, and metadata).
- aipass-bert-detector — A transformer-based model optimized for short-form content (social posts, comments).
(When selecting a model, balance accuracy vs latency and cost. Use detector-lite for high-volume screening and ensemble or v1 for final verification.)
Get Your API Key
Create and manage API keys in the developer dashboard: https://aipass.one/panel/developer.html → API Keys section. Use environment variables (AIPASS_API_KEY) in deployments and never hard-code keys into source.
Related Apps
Explore the app details and UI for this skill: https://aipass.one/apps/ai-detector
Skill File
id: aipass_text_detector
name: AI Pass Text Detector
version: "1.0"
description: Detects whether text was written by AI or a human using the AI Pass AI Text Detector API.
author: AI Assistant
license: MIT
auth:
type: apiKey
name: Authorization
in: header
env: AIPASS_API_KEY
scheme: Bearer
endpoints:
detect_text:
summary: Analyze text and return whether it was written by AI or a human
description: |
Sends the provided text to AI Pass chat/completions endpoint and requests a classification
response indicating whether the text was written by an AI or a human. The skill expects the
AI Pass API key to be available in the environment variable $AIPASS_API_KEY and will include it
as a Bearer token in the Authorization header.
method: POST
url: https://aipass.one/apikey/v1/chat/completions
headers:
Content-Type: application/json
Authorization: "Bearer $AIPASS_API_KEY"
request:
body:
type: object
required:
- model
- messages
properties:
model:
type: string
description: AI Pass model to use for detection. Default may be provider-specific.
default: aipass-detector-001
messages:
type: array
description: Chat-style messages. The user content should be the text to analyze.
items:
type: object
properties:
role:
type: string
content:
type: string
input:
- name: text
in: body
required: true
schema:
type: string
description: The text to analyze for AI/human authorship.
- name: model
in: body
required: false
schema:
type: string
description: Optional model name to send to the API. Defaults to provider's detector model.
build_request:
# Template for request body (replace placeholders as the agent runtime requires)
body:
model: "{{model | default('aipass-detector-001')}}"
messages:
- role: system
content: |
You are an assistant that classifies whether a given text was authored by a human or by an AI.
Provide output in JSON with these fields: label (one of "ai", "human", "uncertain"),
confidence (0.0-1.0), and an explanation string.
- role: user
content: "{{text}}"
response:
type: object
description: The raw API response. The detector's result is typically in choices[0].message.content as JSON.
mappings:
- name: raw
path: "$"
description: Full API response
- name: result_text
path: "choices[0].message.content"
description: Raw string response from detector (recommended: JSON with label/confidence/explanation)
- name: label
path: "choices[0].message.content | json_path('$.label')"
description: Parsed label (ai|human|uncertain) if the model returned JSON
- name: confidence
path: "choices[0].message.content | json_path('$.confidence')"
description: Parsed confidence (0.0-1.0) if present
- name: explanation
path: "choices[0].message.content | json_path('$.explanation')"
description: Parsed explanation string if present
errors:
- code: 401
description: Unauthorized — the API key is missing or invalid. Ensure $AIPASS_API_KEY is set.
- code: 400
description: Bad request — verify the provided text and request body schema.
- code: default
description: Unexpected error returned by the AI Pass API.
examples:
- name: Simple detection
input:
text: "The quick brown fox jumps over the lazy dog."
request:
model: aipass-detector-001
note: |
Example runtime flow:
1) Build the POST body using the template in build_request.
2) Send to https://aipass.one/apikey/v1/chat/completions with Authorization: Bearer $AIPASS_API_KEY.
3) Parse choices[0].message.content as JSON to extract label, confidence, and explanation.
integration_notes:
- Ensure the environment variable AIPASS_API_KEY is set in the agent/runtime environment prior to calling this skill.
- The skill expects the detector to return structured JSON inside choices[0].message.content. If the provider returns free text,
the agent should parse or fallback to returning the raw result_text.
- Adjust the default model name if AI Pass publishes a specific detector model name.