Back to skill

Security audit

agentic-route

Security checks for vulnerabilities and agentic risk

Overview

This skill is aimed at network routing operations, but it exposes disruptive routing, bypass, interception, packet-capture, database, REST API, and MCP actions without enough scoping or safety controls.

Review this carefully before installing. Only use it in an environment where you intentionally want an agent to inspect and modify DNS/routing behavior, and prefer a verified absolute executable path, a pinned trusted install, least-privileged execution, explicit confirmations for mutating commands, and documented rollback procedures.

Vulnerability Patterns
  • Tool Hijacking and SpoofingModifies or replaces tools so legitimate-looking calls execute attacker logic
  • 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 (1)

T07 · Tool Hijacking and Spoofing

Warning
Location
SKILL.md:46
Finding
PATH-Based Executable Spoofing in MCP Configuration## Vulnerability Details **File Location**: `SKILL.md`, lines 46-51 **Vulnerability Type**: Unverified PATH-resolved executable invocation **Risk Level**: Medium The documented MCP configuration invokes `agentic-route` using a bare executable name: ```json { "mcpServers": { "agentic-route": { "command": "agentic-route", "args": ["mcp"] } } } ``` ### Technical Analysis The MCP host resolves the `agentic-route` command through the process environment's `PATH`. The project contains only `SKILL.md` and does not include the referenced executable, a trusted installation source, a pinned version, an integrity checksum, or a verified absolute path. Consequently, the MCP configuration cannot ensure that the program it launches is the intended implementation. An attacker who can place or replace an executable named `agentic-route` in a directory searched before the legitimate installation can cause the MCP host to execute attacker-controlled code. This is a tool-spoofing risk rather than evidence that this package itself contains malicious executable code. The actual behavior of the advertised Bash/Rust engine also cannot be audited because its implementation is absent. ### Attack Path 1. A user or administrator copies the documented MCP configuration into an Agent or MCP host configuration. 2. An attacker gains write access to a directory that appears early in the MCP host's `PATH`, or influences the environment used to launch that host. 3. The attacker places a malicious executable named `agentic-route` in that directory. 4. The MCP host processes the configuration and resolves the bare `agentic-route` command through `PATH`. 5. The malicious executable is launched with the `mcp` argument. 6. The executable runs with the operating-system privileges and environment access of the MCP host process. ### Impact Assessment Successful exploitation permits arbitrary code execution under the ...[truncated 546 chars]
Remediation
## Remediation Suggestions 1. Distribute the actual `agentic-route` implementation through a documented and authenticated installation channel. 2. Pin the installed release to a reviewed version and publish a cryptographic checksum or signed release artifact. 3. Configure MCP with a verified absolute executable path instead of the bare `agentic-route` command. 4. Ensure the executable and all parent directories are owned by a trusted administrator and are not writable by unprivileged users. 5. Launch the MCP server with a minimal, explicitly defined `PATH` and a dedicated least-privileged service account. 6. Audit the missing implementation before deployment, especially the documented packet-capture, routing-rule modification, encrypted-DNS interception, database-access, REST API, and MCP server functionality. 7. Document the expected executable identity, installation location, permissions, version, and integrity-verification procedure so operators can validate it before activation.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The skill documents commands that can alter live routing behavior, remove upstreams, and force bypass/failover without any warning, confirmation step, or discussion of network impact. In an agent-facing skill, this is dangerous because an autonomous system may execute these actions directly, causing outages, traffic interception changes, or loss of name-resolution availability.

Static analysis

No suspicious patterns detected.