Security audit
Security checks for vulnerabilities and agentic risk
This documentation-only skill uses the UnSearch API for web search and scraping as advertised, but users should treat all submitted queries and URLs as data sent to a third party.
Install only if you are comfortable sending search queries, fact-check claims, target URLs, and any extracted page content to UnSearch. Do not use it with secrets, internal-only URLs, private documents, or regulated data unless your policy allows it; use the documented zero-retention header for sensitive-but-approved queries and protect the UNSEARCH_API_KEY like any other credential.
The README promotes web search, URL extraction, and deep research features but does not disclose that user prompts, search terms, and supplied URLs may be transmitted to a third-party service. This can mislead operators into sending sensitive internal queries or private links off-platform without informed consent, creating privacy, compliance, and data-handling risk.
The skill description promotes web search, scraping, and research through a third-party API but does not clearly warn up front that user queries and provided URLs are transmitted off-platform and may cause remote sites to be fetched and processed. This creates a real data-handling and privacy risk because users may supply sensitive prompts, internal URLs, or proprietary targets without realizing they are being sent externally.
Declaring a third-party API base URL and required API-key authentication establishes that the skill depends on external network transmission. In this skill context, that is expected functionality, but it is still a real security-relevant behavior that should be disclosed because user inputs and target URLs leave the local environment.
## API Endpoints
**Base URL:** `https://api.unsearch.dev/api/v1`
All requests require header: `X-API-Key: $UNSEARCH_API_KEY`
The examples demonstrate full-page scraping and extraction workflows without a clear warning that remote page contents will be collected, transmitted, and potentially retained by the external provider. This increases the chance of inadvertent exfiltration of sensitive page contents or use of the tool on internal or private resources without informed consent.
This is a duplicate manifestation of the same external transmission behavior at the concrete search example URL. The context makes it somewhat expected, but the danger remains because the example normalizes sending arbitrary search text and possibly scraped content to a third-party service without nearby warnings.
Search the web with optional content scraping.
curl -X POST "https://api.unsearch.dev/api/v1/search" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
This is a duplicate manifestation of the same external transmission behavior at the concrete search example URL. The context makes it somewhat expected, but the danger remains because the example normalizes sending arbitrary search text and possibly scraped content to a third-party service without nearby warnings.
Search the web with optional content scraping.
curl -X POST "https://api.unsearch.dev/api/v1/search" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
This line again evidences transmission to the external UnSearch agent-search endpoint. In a research skill this behavior is functionally necessary, but absent strong privacy warnings it can still lead to inadvertent disclosure of sensitive prompts and research topics.
AI-optimized search with optional answer generation.
curl -X POST "https://api.unsearch.dev/api/v1/agent/search" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
This line again evidences transmission to the external UnSearch agent-search endpoint. In a research skill this behavior is functionally necessary, but absent strong privacy warnings it can still lead to inadvertent disclosure of sensitive prompts and research topics.
AI-optimized search with optional answer generation.
curl -X POST "https://api.unsearch.dev/api/v1/agent/search" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
This duplicate URL-level finding corresponds to the extract endpoint, which is particularly sensitive because the submitted URLs themselves may grant access to non-public resources or reveal internal systems. The skill context increases risk compared with ordinary web search because extraction invites users to submit direct targets for remote fetching.
Extract content from specific URLs.
curl -X POST "https://api.unsearch.dev/api/v1/agent/extract" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
This duplicate URL-level finding corresponds to the extract endpoint, which is particularly sensitive because the submitted URLs themselves may grant access to non-public resources or reveal internal systems. The skill context increases risk compared with ordinary web search because extraction invites users to submit direct targets for remote fetching.
Extract content from specific URLs.
curl -X POST "https://api.unsearch.dev/api/v1/agent/extract" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
Deep research submits the user's query to an external service that then gathers and synthesizes multi-source content. This can expose sensitive investigative topics, strategic planning questions, or proprietary hypotheses to a third-party processor even if the feature is operationally intended.
Multi-source research with AI synthesis.
curl -X POST "https://api.unsearch.dev/api/v1/agent/research" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
This duplicate endpoint finding again reflects external claim-verification traffic to UnSearch. The feature is legitimate, but without explicit safeguards it may be used on confidential claims or incident narratives that should not be shared with a third party.
Verify claims against multiple sources.
curl -X POST "https://api.unsearch.dev/api/v1/verify/claim" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
This duplicate endpoint finding again reflects external claim-verification traffic to UnSearch. The feature is legitimate, but without explicit safeguards it may be used on confidential claims or incident narratives that should not be shared with a third party.
Verify claims against multiple sources.
curl -X POST "https://api.unsearch.dev/api/v1/verify/claim" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
The Python example embeds the external API base URL and provides ready-to-use code that transmits caller-supplied data off-platform. This is not malicious, but it operationalizes external transmission without built-in consent prompts, validation, or warnings about sending sensitive input.
import os
API_KEY = os.environ["UNSEARCH_API_KEY"]
BASE_URL = "https://api.unsearch.dev/api/v1"
async def search(query: str, scrape: bool = False):
async with httpx.AsyncClient() as client:
The JavaScript example similarly packages external transmission behavior into convenient client code without emphasizing privacy implications. In agent contexts, such snippets are often copied directly into production workflows, increasing the likelihood of silent third-party disclosure of user inputs.
const API_KEY = process.env.UNSEARCH_API_KEY;
const BASE_URL = "https://api.unsearch.dev/api/v1";
async function search(query, scrapeContent = false) {
const response = await fetch(`${BASE_URL}/search`, {
This duplicate endpoint finding appears in the privacy-mode example; despite the privacy framing, data still leaves the environment and is processed remotely. The context can be misleading because users may interpret zero-retention as eliminating all security concerns when it only narrows one aspect of risk.
For sensitive queries, enable zero-retention:
curl -X POST "https://api.unsearch.dev/api/v1/search" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "X-Zero-Retention: true" \
-H "Content-Type: application/json" \
This duplicate endpoint finding appears in the privacy-mode example; despite the privacy framing, data still leaves the environment and is processed remotely. The context can be misleading because users may interpret zero-retention as eliminating all security concerns when it only narrows one aspect of risk.
For sensitive queries, enable zero-retention:
curl -X POST "https://api.unsearch.dev/api/v1/search" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "X-Zero-Retention: true" \
-H "Content-Type: application/json" \
No suspicious patterns detected.