Back to skill

Security audit

site-publisher

Security checks for vulnerabilities and agentic risk

Overview

This skill needs review because its server can expose unrelated local files, and crafted packaging paths can execute commands.

Prefer a patched version that confines reads to the website directory, binds explicitly to localhost, and invokes tar without a shell. Until fixed, avoid exposing the server or supplying untrusted packaging paths. The standalone packaging command produces a folder rather than a ZIP; inspect its contents before sharing. All bundled files were reviewed, but external accessibility depends on your network configuration.

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

Error
Location
.
Finding

Unauthenticated HTTP requests can read files outside the selected site directory

Content
View full analysis

Impact

Attackers can retrieve arbitrary readable files at known or guessed paths, potentially including application configuration and credentials. Both implementations call server.listen(PORT) without a loopback host, allowing non-loopback connections where network policy permits.

Vulnerability Patterns
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Static file serving matches the declared purpose. However, path.join(DIR, req.url) normalizes parent-directory segments without checking that the resulting path remains inside DIR. Consequently, crafted requests can expose files unrelated to the website, exceeding the declared resource scope. Listening without a host restriction can make this exposure reachable beyond localhost, depending on network configuration. The absence of packaging code in this supplied chunk is not itself treated as a mismatch.

Content

No source excerpt is available for this finding.

The documentation promises a ZIP archive, but the implementation only copies files into a directory.

Low
Category
Excessive Agency
Confidence
100% confidence
Finding

The module documentation at L03 and the comment at L14 explicitly describe ZIP creation. Instead, L35–L45 remove '.zip' from the output name, create a directory, and copy the source files into it without producing an archive, contradicting the documented output contract.

Content

No source excerpt is available for this finding.