ClawHub
Back to skill

Security audit

Niri IPC

Security checks for vulnerabilities and agentic risk

Overview

This Niri control skill is coherent, but it gives an agent broad compositor control, raw IPC access, and shell-spawn capability without strong safeguards.

Install only if you want an agent to control your active Niri desktop session. Treat raw IPC, close-window, config reload, output changes, and spawn-sh as high-impact operations; review commands before they run and prefer window IDs or exact matches for destructive actions.

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
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • 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
Findings (7)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding
The code is clearly related to Niri IPC control and is not doing anything unrelated or suspicious. However, the declared description claims a broader capability set than the provided code chunk actually implements. This chunk only supports window/workspace listing and a limited set of actions: focus window, close window, move window to workspace, and focus workspace. It does not implement outputs queries, focused-window queries, command spawning, config reload, or a generic raw IPC interface within this file. Because the evaluation asks whether the description accurately represents what the supplied code chunk actually does, this is a description-behavior mismatch due to materially overbroad declared functionality for the specific code shown.

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill advertises and instructs use of shell commands, environment-dependent IPC access via $NIRI_SOCKET, and command-spawning functionality, but it declares no explicit tool scope or permission boundaries. In an agent setting, that omission weakens policy enforcement and can allow the skill to be invoked with broader shell/environment access than a reviewer or orchestrator expects.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill explicitly documents state-changing and potentially destructive actions such as closing windows, moving windows, reloading configuration, and spawning commands, yet provides no caution, confirmation guidance, or approval boundary. In an autonomous agent context, this increases the risk of accidental disruptive actions, data loss from closing unsaved applications, or abuse of the spawn interfaces to execute arbitrary commands in the user's session.

subprocess module call

Medium
Category
Dangerous Code Execution
Content
if ns.cmd == "event-stream":
        # Stream mode: do not attempt to parse; forward lines.
        cmd = ["niri", "msg", "--json", "event-stream"]
        p = subprocess.Popen(cmd, text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
        assert p.stdout is not None
        count = 0
        try:
Confidence
70% confidence
Finding
subprocess module calls execute external commands. Without careful input validation, this enables command injection.

subprocess module call

Medium
Category
Dangerous Code Execution
Content
def run_json(cmd: list[str]) -> Any:
    p = subprocess.run(cmd, text=True, capture_output=True)
    if p.stderr.strip():
        print(p.stderr.rstrip(), file=sys.stderr)
    if p.returncode != 0:
Confidence
70% confidence
Finding
subprocess module calls execute external commands. Without careful input validation, this enables command injection.

subprocess module call

Medium
Category
Dangerous Code Execution
Content
def run_json(cmd: list[str]) -> Any:
    p = subprocess.run(cmd, text=True, capture_output=True)
    if p.stderr.strip():
        print(p.stderr.rstrip(), file=sys.stderr)
    if p.returncode != 0:
Confidence
70% confidence
Finding
subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The `close` command performs a destructive action immediately after a potentially broad substring match, with no confirmation prompt, dry-run mode, or stronger targeting requirement. In an agent-controlled context, ambiguous matches or prompt manipulation could cause unintended closure of the wrong window, leading to user disruption or data loss in unsaved applications.

Static analysis

No suspicious patterns detected.