Security audit

Office Productivity

Security checks for vulnerabilities and agentic risk

Overview

The skill is not malicious, but it broadly enables email, calendar, and batch document actions without clear safeguards, so it should be reviewed before installation.

Install only if you are comfortable with an office automation skill that may be invoked broadly. For email, calendar, or batch file work, require explicit confirmation of accounts, recipients, folders, files, output paths, and overwrite behavior before letting the agent act.

A.I.G

Vulnerability Patterns

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
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (7)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 90% confidence
Finding: The skill advertises file-generating and file-modifying capabilities but does not declare any explicit tool scope or permissions boundaries. This creates ambiguity about what filesystem actions are authorized and can enable overbroad writes or unsafe execution paths when the agent selects tools automatically.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The trigger language is extremely broad, covering effectively any office-efficiency request, which increases the chance of unintended activation for sensitive tasks. Overbroad activation can cause the agent to invoke email, calendar, or file-processing behaviors in contexts where the user did not clearly consent to those actions.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill supports batch processing and file-modifying document/PDF operations without warning users that files may be altered, overwritten, merged, split, or transformed at scale. In this context, missing warnings increase the risk of accidental data loss, corruption, or unintended processing of sensitive documents.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill describes reading/sending email and accessing calendar data but omits a clear warning that these actions may access sensitive communications, contacts, and scheduling information or send outbound messages on the user's behalf. In an office-productivity context, these capabilities are especially dangerous because they interact with highly sensitive enterprise data and external communication channels.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: This code saves extracted PDF text to a user-specified output file, which can expose sensitive document contents on disk. While it prints a completion message after writing, there is no prior user-facing warning, comment, or docstring disclosing that document contents will be persisted to a text file.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The CLI documentation at L76 advertises 1-based page selections, and the parser at L90-L94 already converts them to zero-based indices. However, split_pdf at L24-L25 subtracts 1 again, causing an off-by-one mismatch between the documented behavior and the actual exported pages.

Natural-Language Policy Violations

Low

Confidence: 95% confidence
Finding: The module docstring is written in Chinese, and the CLI description/help and status messages are also hardcoded in Chinese throughout the file. This creates a language/locale constraint without offering the user a choice or documenting that the tool is intended only for a Chinese-speaking audience.

Static analysis

No suspicious patterns detected.