Back to skill

Security audit

OmniSync_Standard

Security checks for vulnerabilities and agentic risk

Overview

OmniSync does not show exfiltration, persistence, or command execution, but its sync output can be silently altered and its integrity checks are misleading enough to warrant Review before installation.

Install only if you are comfortable with a review-stage sync tool: it has no observed exfiltration, persistence, or command execution, but it should not be trusted for accurate deltas or integrity-sensitive synchronization until the hidden delta substitution, cursor validation bug, gateway startup bug, and input-size enforcement are fixed.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (3)

T09 · Insecure Skill Coding Practices

Error
Location
mcp_gateway.py:47
Finding

Undefined Type Annotation Prevents MCP Gateway Startup

Content
View full analysis
dict: """Formats a standard JSON-RPC success response.""" return { "jsonrpc": "2.0", "result": { "content": [ { "type": "text", "text": f"🔄 OmniSync Standard Result: [{result.get('changed', False)}]\nDelta: {result.get('delta', 'N/A')}\nCursor: {result.get('cursor', 'N/A')}" } ] }, "id": req_id } def format_rpc_error(self, message: str, req_id: Any) -> dict: ``` ### Technical Analysis The `Any` type is used in method annotations but is not imported from the `typing` module. The module also does not enable deferred annotation evaluation with `from __future__ import annotations`. On the documented Python 3.10+ runtime, annotations are evaluated while the class body is defined. Importing or directly executing `mcp_gateway.py` therefore raises a `NameError` when Python encounters the first reference to `Any`. The gateway fails before it can instantiate `OmniSyncGateway` or begin reading JSON-RPC requests. This is a complete availability failure in the primary documented interface rather than a request-specific exception. ### Attack Path 1. A user, service manager, or agent starts the gateway with the documented command: ```bash python mcp_gateway.py ``` 2. Python evaluates the `OmniSyncGateway` class definition. 3. Evaluation reaches the `req_id: Any` annotation. 4. Because `Any` is undefined, Python raises `NameError`. 5. The process exits before entering the stdin request loop. No malicious input is required to trigger the failure. ### Impact Assessment The MCP gateway is completely unavailable, and all `sync_standard` requests through ...[truncated 244 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
mcp_gateway.py:20
Finding

Configured Input-Size Limit Is Not Enforced, Allowing Resource Exhaustion

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
omnisync_engine.py:51
Finding

Incorrect Cursor Validation Allows Changed Content to Be Reported as Unchanged

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill is presented as a standalone synchronization engine, but the detected behavior suggests it actually exposes an MCP/JSON-RPC gateway and delegates work to other components not disclosed in the description. This is dangerous because concealed remote invocation infrastructure expands the attack surface, may permit unintended tool handling or message routing, and prevents reviewers from accurately assessing the real trust and execution boundaries.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding

The skill metadata declares a harmless, standalone sync utility with zero external dependencies, yet the static analysis indicates code capabilities for network and shell without any declared tool scope or permissions boundary. This mismatch is dangerous because hidden execution or communication capabilities can enable unauthorized command execution or data exfiltration while bypassing user expectations and policy controls.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The method claims to compute a linear diff but actually performs inline security filtering and can replace the true delta with a constant marker. In a sync engine, this creates a data integrity vulnerability: downstream systems may believe synchronization succeeded while receiving falsified output, enabling content loss, desynchronization, or deliberate hiding of updates.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The engine is presented as a synchronization/diff component, but it inspects content for selected shell-command substrings and silently substitutes a fake marker instead of returning the actual delta. This hidden content-dependent behavior breaks integrity and predictability of synchronization, and can be abused to suppress or tamper with legitimate content containing those strings.

Content

No source excerpt is available for this finding.