ComPDF Compress PDF

• Outcome: Review
• Latest audit: Aug 3, 2026 · 6:54 AM UTC
• Version: 1.0.0

Overview

This PDF compression skill is mostly coherent, but it packages broad unrelated ComPDF API instructions and a broad default prompt alongside credential use and third-party document upload.

Review this before installing if you handle confidential PDFs. Use it only for explicit PDF compression, confirm any upload to ComPDF, keep the API key file private, and avoid relying on the bundled non-compression API pages unless the skill is narrowed or split into separate capabilities.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Tool Misuse: Tool Parameter Abuse, Chaining Abuse, Unsafe Defaults
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration

Findings (9)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 87% confidence
• Finding: The skill instructs the agent to read local API key files and use network access, but it declares no permissions or trust boundaries for those capabilities. This creates a hidden capability surface: a caller may invoke what appears to be a simple PDF compression skill while the agent accesses sensitive local files and sends data to an external API, increasing the risk of unintended secret use or exfiltration.

Description-Behavior Mismatch

High

• Confidence: 97% confidence
• Finding: The reference file for a skill named specifically for PDF compression embeds the vendor's near-complete API surface, including unrelated conversion, editing, security, generation, and AI extraction features. In an agent setting, this materially expands the action space and can enable capability confusion or prompt-driven misuse beyond the advertised scope, especially around document exfiltration and destructive or privacy-sensitive processing.

Context-Inappropriate Capability

High

• Confidence: 96% confidence
• Finding: Including AI parsing and extraction documentation inside a compression skill exposes powerful data-extraction capabilities unrelated to the task users expect. This is dangerous because uploaded PDFs may contain sensitive business or personal data, and an agent can be induced to extract, structure, and transmit that data to a third-party service under the guise of compression-related help.

Context-Inappropriate Capability

High

• Confidence: 95% confidence
• Finding: The skill materials document PDF encryption and decryption despite the skill being for compression. Decryption is especially sensitive because it can facilitate removal of document protections and normalize processing of protected files outside the user's intended compression workflow.

Context-Inappropriate Capability

High

• Confidence: 96% confidence
• Finding: Broad conversion and generation APIs dramatically exceed the principle of least privilege for a PDF compression skill. This increases the chance that an agent will transform, export, or repackage user documents into other formats, creating unnecessary data exposure, content drift, and opportunities for misuse or accidental exfiltration.

Context-Inappropriate Capability

Medium

• Confidence: 93% confidence
• Finding: Editing, watermarking, insertion, deletion, comparison, and similar PDF operations are outside the skill's stated compression purpose. While some are less sensitive than decryption or AI extraction, they still create an unnecessary and potentially destructive capability set that could alter user documents unexpectedly.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The default prompt instructs the model to use this skill for a broad class of 'document task[s]' without constraining when PDF compression is actually appropriate. Overly broad invocation language can cause the agent to route unrelated or sensitive document workflows through this skill, leading to unintended data disclosure to the ComPDF API or misuse of a transformation step the user did not explicitly request.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The documentation instructs direct file upload and third-party processing but does not clearly warn that potentially sensitive documents will be sent to ComPDF infrastructure. In the context of an agent skill, this can lead users to unknowingly transmit confidential PDFs, increasing privacy, compliance, and data-governance risk.

Tool Parameter Abuse

High

• Category: Tool Misuse

Content

FLAT_REPLACELZW Replace LZW compression with Flate

RMUNUSEDTARGET Remove unused targets

OPTIMIZEPAGECONTENT Optimize page content

• Confidence: 87% confidence
• Finding: RMFORMCOMMITIMPORTRESETACTION Remove form submit/import/

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.