ClawHub
Back to skill

Security audit

plantuml-render

Security checks for vulnerabilities and agentic risk

Overview

This skill appears to be a simple PlantUML renderer, with quality and setup issues but no evidence of hidden, destructive, persistent, or data-stealing behavior.

Before installing, confirm you have a trusted PlantUML JAR available and understand that the skill will create local files and run Java. Treat PlantUML source and included files as local input you control, and expect the published version to need setup fixes before it works reliably.

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
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (5)

Tp4

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The code largely matches the declared core purpose: it uses a PlantUML JAR to render diagrams. However, there are notable description/behavior gaps. The implementation takes a file path (tmp_file) rather than source code content, and the commented-out code suggests direct source handling was intended but is not active. Also, the function docstring states it returns base64-encoded image data, but it actually returns nothing and only prints the output path. While format selection exists in the function via fmt, the provided CLI entry point does not expose alternative formats such as SVG/PDF. These differences are material enough to count as a mismatch, though the primary purpose remains aligned.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill explicitly instructs use of a Python script and java -jar, which means it can invoke shell/subprocess execution, but it declares no permissions or allowed-tools scope. That omission weakens least-privilege controls and can cause the agent to run code-capable actions without clear policy boundaries or user awareness.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill tells the agent to write files and execute a local JAR through Python/Java, but gives no warning about local file modification or subprocess execution. In an agent environment, undisclosed file writes and command execution increase the risk of unsafe operation, especially if user-supplied PlantUML content or file paths are passed through without explicit consent and bounds.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The function documentation explicitly states 'Returns the base64-encoded image data.' However, the implementation invokes the PlantUML JAR, derives an output file path, and prints that path; it does not read the generated file, encode it, or return any value. This is an active contradiction between documented intent and actual behavior.

Content

No source excerpt is available for this finding.

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/plantuml.py (reported line 20)May include surrounding context.

python
if out_dir:
        cmd += ["-o", out_dir]
    cmd.append(tmp_file)
    subprocess.run(cmd, check=True)
    img_path = output_path or os.path.join(os.path.dirname(tmp_file), os.path.splitext(os.path.basename(tmp_file))[0]+f".{fmt}")
    print(f"image path {img_path}")