ClawHub
Back to skill

Security audit

document-ocr-agent

Security checks for vulnerabilities and agentic risk

Overview

This OCR skill is a coherent remote document-processing integration, but users should treat submitted documents as sensitive data sent to AgentPMT and Google Document AI.

Install only if you are comfortable sending selected documents to AgentPMT/Google Document AI for OCR. Avoid submitting passports, driver's licenses, tax forms, bank statements, medical records, or contracts unless you are authorized to share them, and prefer redacted/minimized inputs. Use pinned or trusted install sources for the related setup skills when possible.

A.I.G

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:236
Finding

Unpinned Third-Party Skill Installation Permits Supply-Chain Code Execution

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 236-260
Vulnerability Type: Unpinned third-party dependencies installed through npx
Risk Level: Medium

The affected instructions include:

markdown
Core AgentPMT setup skills:
- What AgentPMT is: ../what-is-agentpmt
  - ClawHub page: https://clawhub.ai/agentpmt/what-is-agentpmt
  - OpenClaw install: `openclaw skills install what-is-agentpmt`
  - skills.sh install: `npx skills add AgentPMT/agent-skills --skill what-is-agentpmt`
- AgentPMT account MCP/REST setup: ../agentpmt-account-mcp-rest-api-setup
  - ClawHub page: https://clawhub.ai/agentpmt/agentpmt-account-mcp-rest-api-setup
  - OpenClaw install: `openclaw skills install agentpmt-account-mcp-rest-api-setup`
  - skills.sh install: `npx skills add AgentPMT/agent-skills --skill agentpmt-account-mcp-rest-api-setup`

skills.sh install script:

```bash
npx skills add AgentPMT/agent-skills --skill what-is-agentpmt
npx skills add AgentPMT/agent-skills --skill agentpmt-account-mcp-rest-api-setup
text

### Technical Analysis

The Skill instructs users to run `npx` and install skills from mutable package and repository references. The commands do not pin an exact package version, immutable repository commit, or cryptographic integrity digest. Consequently, the code retrieved and executed during installation can differ from the content available when this Skill was audited.

`npx` may download and execute package code on the local system. Likewise, installing a named remote skill without an immutable revision delegates trust to the package registry, repository, publisher accounts, and their future releases. If any part of this supply chain is compromised, malicious setup code could execute under the invoking user's account.

The Skill also advises users to reinstall it when its freshness period expires. Reinstalling from an unpinned source increases exposure to later changes that were not part of the reviewed artifact.

No evidence shows th
...[truncated 1449 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin the installer package to an exact, reviewed version instead of allowing npx to resolve the latest release.
  2. Pin each external skill to an immutable repository commit or signed release artifact.
  3. Publish and verify cryptographic checksums or signatures before installation.
  4. Use a trusted package registry with provenance verification and package-lock enforcement.
  5. Prefer installation modes that do not execute lifecycle scripts unless those scripts have been reviewed and are strictly necessary.
  6. Document the expected publisher, package version, commit identifier, and integrity digest so users can verify the retrieved artifact.
  7. Run installation with a dedicated, least-privileged account and without unrelated credentials in the environment.
  8. Require security review before updating pinned versions rather than automatically trusting newly published content.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (14)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill encourages uploading receipts, invoices, IDs, tax forms, passports, and other high-sensitivity documents to a remote OCR service, but it does not clearly warn users that this content may contain personal, financial, or regulated data that will leave the local environment. That omission can lead to unintentional exfiltration of sensitive information and defeats informed consent around external processing.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The activation keywords are broad enough to match common OCR and document-processing requests, which can cause over-triggering of this remote skill in situations where the user did not intend third-party document transmission. In this skill's context, that matters because the tool handles potentially sensitive financial, tax, identity, and medical documents and sends them to an external service.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

This skill is explicitly designed to transmit user-supplied documents to external AgentPMT endpoints for OCR processing, which constitutes real external data transmission. In this context, the danger is elevated because the supported use cases include passports, driver's licenses, W-2s, W-9s, bank statements, receipts, medical records, and contracts, all of which may contain highly sensitive data.

Content

Scanner excerpt · SKILL.md (reported line 346)May include surrounding context.

md
- What AgentPMT is: ../what-is-agentpmt (ClawHub: `what-is-agentpmt`, page: https://clawhub.ai/agentpmt/what-is-agentpmt; skills.sh: `npx skills add AgentPMT/agent-skills --skill what-is-agentpmt`)
- AgentPMT account MCP/REST setup: ../agentpmt-account-mcp-rest-api-setup (ClawHub: `agentpmt-account-mcp-rest-api-setup`, page: https://clawhub.ai/agentpmt/agentpmt-account-mcp-rest-api-setup; skills.sh: `npx skills add AgentPMT/agent-skills --skill agentpmt-account-mcp-rest-api-setup`)
- Marketplace product: https://www.agentpmt.com/marketplace/google-document-ai-ocr
- AgentPMT main MCP server: https://api.agentpmt.com/mcp/
- AgentPMT REST invoke endpoint: https://api.agentpmt.com/products/purchase

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

The REST invocation endpoint confirms that document contents can be sent off-platform to a remote service, making this a true data-exposure risk rather than a hypothetical one. Because the skill markets itself for OCR of financial, tax, and identity documents, the sensitivity of likely payloads makes external transmission materially more dangerous.

Content

Scanner excerpt · SKILL.md (reported line 347)May include surrounding context.

md
- AgentPMT account MCP/REST setup: ../agentpmt-account-mcp-rest-api-setup (ClawHub: `agentpmt-account-mcp-rest-api-setup`, page: https://clawhub.ai/agentpmt/agentpmt-account-mcp-rest-api-setup; skills.sh: `npx skills add AgentPMT/agent-skills --skill agentpmt-account-mcp-rest-api-setup`)
- Marketplace product: https://www.agentpmt.com/marketplace/google-document-ai-ocr
- AgentPMT main MCP server: https://api.agentpmt.com/mcp/
- AgentPMT REST invoke endpoint: https://api.agentpmt.com/products/purchase

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The schema describes sending documents, URLs, and file IDs to a remote OCR service but does not clearly warn users that document contents and references will be transmitted to Google Document AI for third-party processing. Because this skill handles potentially sensitive records such as passports, W-2s, W-9s, invoices, and bank statements, missing disclosure increases the risk of unintended exposure of confidential or regulated data.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill repeatedly documents that callers must provide exactly one of file_urls, file_ids, or content_base64 and that providing multiple sources returns an error. However, both the MCP and REST example call bodies include all three inputs simultaneously, which directly contradicts the documented action contract and can mislead integrators about valid usage.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The schema repeatedly documents content_base64 as base64-encoded file content, but the sample value is ordinary text (Draft marketing copy to check for banned phrases.) rather than encoded document bytes. This actively misrepresents how the action should be used and suggests a different text-analysis behavior than the OCR/document-processing purpose described elsewhere.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.