Security audit
Security checks for vulnerabilities and agentic risk
This skill is a straightforward Upstage document-parsing helper that sends chosen documents to Upstage’s API, with privacy considerations users should understand before use.
Install only if you are comfortable sending selected documents to Upstage for processing. Avoid uploading confidential, regulated, or highly sensitive files unless your organization permits that use of Upstage and you understand the remote retention behavior for async results.
The skill clearly uploads user-supplied documents to the Upstage API, but the user-facing guidance does not prominently warn that document contents leave the local environment and are sent to a third-party service. This can cause unintended disclosure of sensitive files if users or calling agents assume processing is local.
This code performs an external HTTP POST that transmits the full document to a third-party API for parsing. In the context of document processing, uploaded files may contain confidential, regulated, or proprietary content, so undisclosed external transmission creates a real privacy and data governance risk.
import requests
with open("report.pdf", "rb") as f:
response = requests.post(
"https://api.upstage.ai/v1/document-digitization",
headers={"Authorization": f"Bearer {os.environ['UPSTAGE_API_KEY']}"},
files={"document": f},
The hardcoded external endpoint confirms that document content is sent to Upstage infrastructure rather than processed locally. That is expected for this integration, but it is still security-relevant because users may expose sensitive document contents to an external processor if the skill does not clearly disclose that behavior.
with open("report.pdf", "rb") as f:
response = requests.post(
"https://api.upstage.ai/v1/document-digitization",
headers={"Authorization": f"Bearer {os.environ['UPSTAGE_API_KEY']}"},
files={"document": f},
data={"model": "document-parse", "output_formats": "['markdown']"}
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.upstage.ai/v1/document-digitization/async" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@large.pdf" \
-F "model=document-parse" \
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.upstage.ai/v1/document-digitization/async" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@large.pdf" \
-F "model=document-parse" \
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.upstage.ai/v1/document-digitization/async" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@large.pdf" \
-F "model=document-parse" \
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.upstage.ai/v1/document-digitization/async" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@large.pdf" \
-F "model=document-parse" \
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.upstage.ai/v1/document-digitization/async" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@large.pdf" \
-F "model=document-parse" \
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
# 1. Submit async request
with open("large.pdf", "rb") as f:
r = requests.post(
"https://api.upstage.ai/v1/document-digitization/async",
headers={"Authorization": f"Bearer {api_key}"},
files={"document": f},
The examples explicitly upload local files such as PDFs and invoices to a third-party API but do not warn users that document contents will leave the local environment. Because this skill is specifically about document parsing, users may submit sensitive business, financial, or personal documents and overlook the privacy, compliance, and data handling implications.
This example sends a local document to an external HTTPS endpoint, which is expected behavior for a cloud parsing API, but it still creates a real data exfiltration path for any file the user supplies. In the context of documentation, the issue is not hidden malware but the absence of disclosure and safeguards around transmitting potentially sensitive document contents off-device.
curl -X POST "https://api.upstage.ai/v1/document-digitization" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@report.pdf" \
-F "model=document-parse" \
This command uploads a local file named complex.pdf to the Upstage API, so any embedded sensitive text, tables, or images are transmitted externally. Although this is consistent with the product's purpose, users are not warned about the resulting privacy and compliance risks, which makes the example unsafe in practice for unaware users.
curl -X POST "https://api.upstage.ai/v1/document-digitization" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@complex.pdf" \
-F "model=document-parse" \
The scanned-document OCR example transmits scan.pdf to an external service, which may include highly sensitive scanned records like IDs, contracts, or medical paperwork. Since the example lacks any warning about off-system processing, it could cause unintentional exposure of sensitive scanned content.
curl -X POST "https://api.upstage.ai/v1/document-digitization" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@scan.pdf" \
-F "model=document-parse" \
This example uploads invoice.pdf to a remote API and additionally requests base64-encoded table extraction, which may increase the amount of structured sensitive financial data returned and handled. Invoices commonly contain PII, account details, and billing information, so the lack of explicit privacy warning makes this more dangerous in context.
curl -X POST "https://api.upstage.ai/v1/document-digitization" \
-H "Authorization: Bearer $UPSTAGE_API_KEY" \
-F "document=@invoice.pdf" \
-F "model=document-parse" \
The description embeds Korean example invocations alongside English ones, which introduces a language-specific interaction pattern without stating whether the skill is multilingual, user-selectable, or region-specific. Under the policy, language constraints or preferences should be optional or clearly justified rather than implicitly imposed.
No suspicious patterns detected.