Security audit

Omni Tool

Security checks for vulnerabilities and agentic risk

Overview

This skill is mainly a local logging tool, but its metadata advertises unrelated bookmark/Chrome functionality while the script persistently stores raw user input in plaintext.

Review this before installing if you expected a bookmark or Chrome tool. Use it only for non-sensitive notes or task logs, because anything passed to its commands may be saved in plaintext under ~/.local/share/omni-tool and later shown by recent/search-style commands.

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
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
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)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 98% confidence
Finding: The skill metadata and description advertise a JavaScript/Chrome bookmark management utility, but the documented behavior is a Bash-based local logging tool with unrelated commands. This kind of description-behavior mismatch is dangerous because it can mislead users and downstream systems into granting trust, selecting the skill for inappropriate tasks, or exposing sensitive inputs that are then silently persisted to local logs.
Content: No source excerpt is available for this finding.

Ssd 3

Medium

Category: Not specified by scanner
Confidence: 99% confidence
Finding: Across multiple commands, the tool persistently stores and later displays user-supplied content in plain language, creating a broad local data exposure surface. Given the manifest claims a browser/bookmarks productivity function, the context makes this more dangerous because users may reasonably submit browsing-related or productivity data without realizing it is being archived and replayed.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 97% confidence
Finding: The script's advertised purpose does not match its implementation: instead of browser/bookmark management, it exposes a generic local logging utility. This deception increases the chance users will provide data under false assumptions, and the mismatch obscures the real behavior of persistent collection and later disclosure of inputs.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 96% confidence
Finding: The export feature aggregates all previously logged plaintext entries into new files without warning that potentially sensitive content will be duplicated and broadened in exposure. Exporting to json, csv, or txt can make data easier to copy, share, ingest, or exfiltrate accidentally, especially since values are not sanitized or filtered.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 98% confidence
Finding: User input passed to commands is written directly into persistent plaintext log files under the user's home directory without clear notice or consent. If a user enters secrets, personal data, URLs, tokens, or work content, that data remains on disk and can later be accessed by other local processes, backups, or anyone with filesystem access.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 98% confidence
Finding: Commands such as convert, analyze, generate, compare, and report are misleading because they do not perform the named actions and instead silently store the full user-provided input. In a productivity skill context, users may paste sensitive content expecting processing, but the tool merely retains it in plaintext, creating an unnecessary privacy and trust risk.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.