lintai docs

Appearance

Rule Reference

SEC730lintai-ai-securitythreat-reviewsecuritystablejsonwarn

MCP config: keylogger exfiltration

MCP configuration command captures and exfiltrates keystrokes or keyboard input

Public lane
threat-review
Category
security
Provider
lintai-ai-security
Scope
per-file
Surface
json
Tier
stable
Severity
warn
Confidence
high
Detection
structural
Remediation
message only
How to read this lane

Explicit malicious, secret-bearing, or spyware-like review.

How to read this category

Strong exploit, secret, or unsafe-execution signal.

Activation Model

Preset Activation

These presets explain where this rule appears in the product experience.

threat-reviewsidecar lane
mcpsurface preset

Lifecycle

Stable Lifecycle Contract

State

stable

Graduation rationale

Checks committed MCP launch paths for explicit keystroke capture utilities or keylogger markers combined with remote transfer behavior.

Deterministic signal basis

JsonSignals command-plus-args analysis over ArtifactKind::McpConfig for explicit keylogger primitives such as `logkeys`, `xinput test`, `evtest`, `showkey`, PowerShell `GetAsyncKeyState`, or inline Python listener markers like `pynput.keyboard.Listener`, combined with remote sinks such as `curl`, `wget`, `scp`, `rsync`, `nc`, or HTTP(S) endpoints.

Malicious corpus
mcp-command-keylogger-exfil
Benign corpus
mcp-safe-basic
structured evidence required remediation reviewed
Canonical note

Structural stable rule positioned as an explicit threat-review control: high-signal malicious, credential-bearing, or spyware-like behavior that stays opt-in rather than shaping the quiet default.

Nearby Signals

Related Rules

SEC301MCP config: shell trampolineMCP configuration shells out through sh -c or bash -cthreat-reviewsecurityjson
SEC305Config: hardcoded auth materialConfiguration embeds static authentication material in a connection or auth valuethreat-reviewsecurityjson
SEC309Config: literal secrets in configConfiguration commits literal secret material in env, auth, or header valuesthreat-reviewsecurityjson
SEC310Config: metadata or private-network hostConfiguration endpoint targets a metadata or private-network host literalthreat-reviewsecurityjson
SEC338MCP config: sensitive host bind mountMCP configuration launches Docker with a bind mount of sensitive host materialthreat-reviewsecurityjson
SEC339MCP config: privileged Docker flagsMCP configuration launches Docker with a host-escape or privileged runtime flagthreat-reviewsecurityjson

Why It Matters ​

Capturing keyboard input and sending it to a remote destination is high-confidence surveillance and credential theft behavior.

What Triggers ​

SEC730 matches MCP commands that invoke explicit keylogger primitives such as logkeys, xinput test, evtest, showkey, PowerShell GetAsyncKeyState, or inline Python listener markers like pynput.keyboard.Listener, and also transmit data to remote sinks such as curl, wget, scp, rsync, nc, or HTTP(S) endpoints.

False Positives ​

Shared committed MCP launch paths should not capture and transmit keyboard input. Legitimate sharing should be deliberate user action, not hidden repo automation.

Remediation ​

Remove keystroke capture and remote transfer behavior from the committed MCP launch path and keep keyboard data local unless the user explicitly chooses to share it.