Back to skill

Security audit

gateway-watchdog

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because scanners report concerns with its administrator installer, background watchdog, and maintenance controls; the scan was partial, so the evidence is incomplete.

Consider the reported risks to administrator task setup, executable selection, and pausing the background watchdog before installing. Skillspector inspected 11 of 12 components, and its tool-misuse analysis was degraded for one installer file; several local references were also unresolved, so coverage is incomplete.

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 (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/install_watchdog_admin.ps1
Finding

Alternative installer bypasses confirmation before persistent task replacement

Content
View full analysis

Impact

Persistent task definitions can be replaced and background watchdog execution enabled without the operator reviewing and approving those changes at this entry point. Existing task configuration is overwritten. Administrator execution is a prerequisite; this is not an unauthenticated privilege-escalation vulnerability.

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (6)

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · CHANGELOG.md (reported line 72)May include surrounding context.

md
Two SkillSpector findings were reviewed and dismissed:

- **AR2 (Anti-Refusal, HIGH)** — the phrase `do NOT judge` in `watchdog.ps1` is a
  comment explaining that the maintenance lock suspends health judgement. It is
  not an instruction to suppress user warnings. The auditor itself recorded this
  as not accepted.

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · scripts/uninstall_watchdog_admin.bat (reported line 25)May include surrounding context.

text
echo [1/3] Removing watchdog task...
schtasks /End /TN "OpenClaw Watchdog" >nul 2>&1
schtasks /Delete /TN "OpenClaw Watchdog" /F >nul 2>&1
if %errorlevel% equ 0 (echo     [OK] Task removed.) else (echo     [!] Task not found or could not be removed.)
echo.

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · scripts/watchdog.ps1 (reported line 290)May include surrounding context.

text
$cycle++

        # --- MAINTENANCE GATE ---
        # If a repair (ours or external) holds a fresh lock, do NOT judge
        # the gateway health or attempt a restart. Just wait.
        if (Test-MaintenanceLock) {
            if ($lastStatus -ne "maintenance") {

The documented pause command does not stop an already-running watchdog.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

Disabling a scheduled task prevents future launches but does not terminate an already-running instance, contrary to the documented promise that the watchdog stops acting. An operator could begin maintenance believing the hidden watchdog is paused while it continues restarting the Gateway, disrupting repairs or availability.

Content

No source excerpt is available for this finding.

The resolver trusts PATH despite explicitly claiming it never does so with elevation.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

If the candidate files are absent, the resolver falls back to Get-Command openclaw, allowing command resolution through PATH despite the stated restriction. An attacker who controls a searched location could supply a malicious executable that receives the wrapper's privileges, including administrator privileges if the wrapper is elevated; the argument allowlist does not authenticate the executable.

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · CHANGELOG.md (reported line 76)May include surrounding context.

md
comment explaining that the maintenance lock suspends health judgement. It is
  not an instruction to suppress user warnings. The auditor itself recorded this
  as not accepted.
- **EA3 (Excessive Agency, LOW)** — matched the string `NOT LIMITED TO` inside the
  MIT-0 warranty disclaimer. Not a capability statement.

### Carried over from 1.1.0