Security audit

Pg Status

Security checks for vulnerabilities and agentic risk

Overview

The status skill is mostly read-only on its main page, but it ships a broad ProxyGate command reference with money-moving, account, seller-admin, tunnel, and job mutation commands that do not fit a status-only skill.

Review before installing. Use this only if you are comfortable with a ProxyGate skill package that includes broad CLI documentation beyond status checks, and require explicit confirmation before any deposit, withdraw, auth change, listing change, proxy request, tunnel, or job action.

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

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (6)

Description-Behavior Mismatch

High

Category: Not specified by scanner
Confidence: 98% confidence
Finding: The skill is explicitly described as a status-checking skill, but the reference file exposes a broad operational CLI including authentication changes, deposits, withdrawals, listing mutation, proxy execution, tunnel exposure, and job actions. In an agent setting, oversized command references can steer the model into performing unintended state-changing actions far outside the declared scope, increasing the chance of unauthorized or destructive behavior.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

High

Category: Not specified by scanner
Confidence: 99% confidence
Finding: This section includes high-impact transactional actions such as deposit, withdraw, wallet generation, and auth changes, none of which are necessary for checking status. Because the skill is triggered by balance and account-status queries, including money-moving commands in this context materially increases the risk that an agent may execute irreversible financial operations when the user only intended read-only inspection.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

High

Category: Not specified by scanner
Confidence: 98% confidence
Finding: The seller/admin section contains mutation operations like create, update, pause, unpause, delete, rotate-key, upload-docs, and header changes, which go well beyond status retrieval. In the context of a status skill, these commands create a confused-deputy risk where an agent may alter marketplace configuration or secrets while handling benign account overview requests.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The trigger description is broad enough to activate on generic phrases like 'check balance', 'my usage', or 'what's my balance' without requiring clear ProxyGate context. This can cause the wrong skill to be invoked for unrelated tasks, leading to unintended command suggestions, confusion, and possible exposure of ProxyGate-specific account or operational actions in inappropriate contexts.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium

Category: Not specified by scanner
Confidence: 94% confidence
Finding: Arbitrary proxy request execution allows the agent to invoke external services, send request bodies, stream responses, and potentially incur costs or trigger side effects unrelated to status checks. Although not necessarily destructive by itself, this materially expands the action surface of a status-oriented skill and could be abused to perform unintended API calls under the guise of account inspection.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: Tunnel exposure, development tooling, and job marketplace actions are unrelated to status retrieval and can change external exposure, workflow state, or financial/job outcomes. Their presence inside a status skill makes the skill more dangerous because the trigger phrases are broad and likely to be used in routine account checks, yet the referenced commands enable materially different and potentially harmful operations.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.