Security audit

MaybeAI Sheet CLI

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent MaybeAI spreadsheet CLI helper, but it needs Review because it enables public/editor sharing and LLM-based workbook profiling without strong confirmation and privacy guardrails.

Install only if you intend to let `mbs` use `MAYBEAI_API_TOKEN` to read, modify, import, export, and share MaybeAI spreadsheets. Treat `workbook metadata` as potentially sending worksheet samples to the service for LLM summarization and caching. Require explicit confirmation before public sharing or editor grants, verify the exact sheet, recipient email, and permission, and prefer private/viewer access when possible.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 83% confidence
Finding: The skill clearly instructs use of local file imports, write operations, and remote URL imports, which means it exercises file read/write and network capabilities without any declared permission model or safety boundary. That increases the chance an agent invokes sensitive filesystem or outbound network actions without explicit user awareness or policy enforcement.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The skill provides commands to make workbooks public and grant viewer/editor access, but it does not require explicit warnings, confirmation steps, or least-privilege guidance before changing visibility. In an agent setting, that can lead to accidental exposure of sensitive spreadsheet data or overbroad sharing initiated from ambiguous user prompts.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The documentation exposes high-impact sharing operations such as making a workbook public or granting editor access without any caution, confirmation guidance, or privacy/access-control warning. In an agent skill context, this increases the chance that an automated agent or user follows the examples verbatim and unintentionally exposes sensitive spreadsheet data to unauthorized parties.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The reference provides direct commands to make a sheet public, grant editor access, and remove access, but it does not explicitly warn about the privacy and integrity consequences of exposing data or granting write permissions. In an agent skill, this omission can cause the agent to carry out high-impact sharing changes from terse user prompts without prompting for confirmation or emphasizing that data may become broadly accessible.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The document explicitly states that non-empty sample rows from each worksheet are sent to an LLM, but it does not prominently warn users about external data transmission, privacy implications, or the possibility that sampled rows may contain sensitive business or personal data. In a workbook-analysis skill, even small worksheet samples can expose confidential information, so silent transmission to an LLM is a meaningful data-handling risk.

Natural-Language Policy Violations

Medium

Confidence: 82% confidence
Finding: The skill says the service generates a Chinese natural-language summary and later reiterates that the summary is generated in Chinese, without indicating user choice or locale negotiation. This is not a classic security flaw, but it can create integrity and usability issues if users misread the output, miss warnings, or rely on summaries they cannot accurately validate.

Static analysis

No suspicious patterns detected.