Security audit

Skill Network Audit: where skills send data

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed network-audit helper that reads a local ClawPhylax ledger and includes user-approved host allow/deny actions scoped to a skill origin.

Before installing, understand that this skill relies on ClawPhylax ledger data that may reveal internal hostnames, service usage, uploads, and credential-read indicators. Treat reports and witness cards as potentially sensitive, and approve any allow/deny or posting action only when you intend to change enforcement or share hostnames.

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
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill is described primarily as a read-only network audit capability, but it also documents commands that change enforcement state by allowing or denying hosts. That mismatch can mislead users or higher-level agents into invoking a skill with side effects they did not expect, creating a permission and trust-boundary problem even if the mutation is presented as optional.

Missing User Warnings

Low

Confidence: 84% confidence
Finding: The markdown explains that the skill reads a local ledger of outbound connections and includes details such as hosts contacted and credential reads, but it does not plainly warn the user that using the skill may reveal sensitive infrastructure or privacy-relevant metadata. Because this is a markdown skill description for a data-sensitive auditing capability, a brief user-facing warning would improve informed use.

Static analysis

No suspicious patterns detected.