Security audit

文档解析工具（专业版）

Security checks for vulnerabilities and agentic risk

Overview

This is a document-parsing skill with mostly expected capabilities, but its handling of sensitive document contents, callbacks, and saved outputs is not scoped clearly enough.

Install only if you are comfortable letting the skill read user-selected documents, install parsing/OCR dependencies, write extracted text and tables to disk, and potentially use API keys or callbacks. Avoid confidential documents unless you control the output directory, disable or review callbacks/webhooks, and know how generated files and audit records will be cleaned up.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Vague Triggers

Medium

Confidence: 82% confidence
Finding: The activation guidance is broad enough to match many ordinary file-handling requests, increasing the chance that an agent invokes a skill with read/write/exec capabilities in contexts where a simpler, safer path would suffice. Over-broad triggering materially expands the attack surface by routing sensitive user documents and file operations through a powerful skill unnecessarily.

Vague Triggers

Low

Confidence: 79% confidence
Finding: The example triggers such as parsing a document structure or extracting text/tables are generic and weakly bounded, which can cause accidental or excessive skill activation. In a skill that advertises read/write/exec tool access, vague trigger phrases increase the likelihood of unnecessary exposure of local files and derived data.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The skill accepts document input and a callback URL but does not clearly warn that parsed contents may include sensitive data and that callbacks may transmit data to external endpoints. This omission is dangerous because users or orchestrators may unknowingly send confidential document content or metadata off-host without informed consent or validation.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill describes writing output files locally and storing results for audit/reporting, but it does not clearly warn that parsed documents and extracted text may persist on disk. This creates a real confidentiality risk, especially for sensitive documents, because local artifacts and audit records can outlive the original task and be accessed by other users, processes, or backups.

Static analysis

No suspicious patterns detected.