ClawHub
Back to skill

Security audit

GStack Review Stack

Security checks for vulnerabilities and agentic risk

Overview

This skill is a review workflow adapter that is mostly transparent and proportionate, with some disclosed use of local or public gstack reference material.

Before installing, know that this skill can steer review and ship/no-ship recommendations and may read a local gstack clone or inspect the public gstack repository if you ask for upstream behavior. Do not let it publish, unpublish, delete, or submit anything unless you explicitly approve that action after reviewing the recommendation.

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
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The declared description says this skill should be used to conduct various gstack-style reviews and decision processes. However, the supplied code does not implement any review, critique, QA gate, release gate, or portfolio decision logic. Its primary function is simply to locate a local gstack clone by checking specific directories for README.md and plan-ceo-review/SKILL.md, then list skill definition files. This is a materially different primary purpose from the declared behavior, so it should be flagged as a mismatch.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 88)May include surrounding context.

md
If `/home/z/Github/gstack` exists, use it as source material for deeper review. Read only the specific upstream `SKILL.md` needed for the current mode.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest description says to trigger for phrases like "CEO review," "design critique," and "QA gate," which are broad natural-language requests that could arise in many contexts without clearly indicating this specific skill should run. The file does not provide exclusion conditions or negative examples to bound when these triggers should not activate.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill instructs fetching or inspecting external GitHub content when a local clone is missing, which expands execution beyond the local workspace and introduces unpinned, mutable third-party content into the review flow. This can lead to prompt injection, trust of attacker-controlled updates, or data exposure through unnecessary network access, especially in an agent setting where fetched content may influence later decisions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The phraseology in this file is broad enough to be triggered by generic requests such as release readiness or publishing decisions, which can cause the skill to activate outside narrowly intended scenarios. In an agent setting, unintended invocation can steer workflows, produce authoritative ship/no-ship recommendations, or influence publishing actions when the user did not specifically request this review mode.

Content

No source excerpt is available for this finding.