Agentinel / MCP security scan · v0.1.0

kubernetes

FAIL · 1 high
1 server 6 tools from published tool specs

Findings

2
Total
0
Critical
1
High
0
Medium
1 high1 info

By OWASP Agentic category

ASI02Tool Misuse and Exploitation1
ASI05Unexpected Code Execution1

Lethal trifecta

Rule of Two
A
Untrusted inputreads content an attacker can write
0 tools
B
Sensitive datareaches private data or systems
3 tools
C
External actionchanges state or sends data out
3 tools
Trifecta broken. 2 of 3 axes present, so an injection can't complete an exfiltration path on its own.

Capability matrix · Rule of Two

Server / tool Auntrusted input Bsensitive data Cexternal action
kubernetes6 tools
kubectl_apply
kubectl_delete
kubectl_generic
kubectl_get
kubectl_logs
kubectl_scale

Findings · most severe first

HIGH

Unsafe code or command execution surface

ASI05
kubernetes:kubectl_generic

Tool 'kubectl_generic' exposes code or command execution to the agent, so an injected instruction could run commands on the host.

matched: execute any
Sandbox execution, drop privileges, allow-list commands, and require human approval for any code-execution tool.
ASI05 · Unexpected Code Executionconfidence 70%
INFO

Two of three lethal-trifecta axes present

ASI02
server:kubernetes

This server's toolset already holds two trifecta axes (B:sensitive-access, C:external-comms). Adding a tool with the third axis — or connecting it alongside another server that provides it — would complete the lethal trifecta.

axes present: B:sensitive-access, C:external-comms
Break the trifecta: remove one axis, split capabilities across isolated agents, or require human approval for the state-changing / external-comms step.
ASI02 · Tool Misuse and Exploitationconfidence 60%