Back to skill

Security audit

open-perplexity

Security checks for vulnerabilities and agentic risk

Overview

This skill is a small disclosed browser launcher for Perplexity in Brave, with minor safety and quality caveats but no evidence of hidden or malicious behavior.

Installing this should mainly let the agent open Perplexity in Brave. Review the broad trigger phrase and the Windows shell=True implementation if you want tighter activation and safer process launching, but the inspected package does not request credentials, persist data, or perform hidden system changes.

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
Vulnerability Patterns
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% confidence
Finding

This duplicate finding points to the same Windows shell-based subprocess invocation and represents the same underlying issue: unnecessary shell execution for opening a browser. In an agent skill context, launching local processes is expected, but invoking them via the shell broadens attack surface without any functional need.

Content

Scanner excerpt · scripts/open_perplexity.py (reported line 9)May include surrounding context.

python
system = platform.system()

if system == "Windows":
    subprocess.Popen(["start", "brave", url], shell=True)
elif system == "Darwin":
    subprocess.Popen(["open", "-a", "Brave Browser", url])
else:

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% confidence
Finding

This duplicate finding points to the same Windows shell-based subprocess invocation and represents the same underlying issue: unnecessary shell execution for opening a browser. In an agent skill context, launching local processes is expected, but invoking them via the shell broadens attack surface without any functional need.

Content

Scanner excerpt · scripts/open_perplexity.py (reported line 9)May include surrounding context.

python
system = platform.system()

if system == "Windows":
    subprocess.Popen(["start", "brave", url], shell=True)
elif system == "Darwin":
    subprocess.Popen(["open", "-a", "Brave Browser", url])
else:

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding

The skill description indicates shell-based browser launching across operating systems, but the manifest does not declare any tool scope or allowed tools. This creates a capability/permission mismatch: an agent or reviewer cannot accurately determine what the skill is allowed to do, and shell access is especially sensitive because it can be repurposed beyond simply opening a URL.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Several trigger phrases are broad or ambiguous, especially "open ai search," which can overlap with unrelated user intent and cause accidental invocation. Even for a benign skill, overbroad activation increases the chance of unintended browser launches and reduces user control over when the skill executes.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
87% confidence
Finding

The Windows branch launches a subprocess with shell=True, which invokes the system shell and increases the risk of command interpretation issues. In this specific file the URL is hardcoded, so there is no immediate user-controlled injection path, but using shell=True for process launch is still an unsafe pattern that can become exploitable if the URL, browser name, or environment is later made dynamic.

Content

Scanner excerpt · scripts/open_perplexity.py (reported line 9)May include surrounding context.

python
system = platform.system()

if system == "Windows":
    subprocess.Popen(["start", "brave", url], shell=True)
elif system == "Darwin":
    subprocess.Popen(["open", "-a", "Brave Browser", url])
else:

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/open_perplexity.py (reported line 11)May include surrounding context.

python
if system == "Windows":
    subprocess.Popen(["start", "brave", url], shell=True)
elif system == "Darwin":
    subprocess.Popen(["open", "-a", "Brave Browser", url])
else:
    subprocess.Popen(["brave-browser", url])

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/open_perplexity.py (reported line 13)May include surrounding context.

python
elif system == "Darwin":
    subprocess.Popen(["open", "-a", "Brave Browser", url])
else:
    subprocess.Popen(["brave-browser", url])

print("✅ Perplexity opening in Brave!")