Security audit

PowerfulHarp2041

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly does test-effort estimation, but it also ships an unrelated packaging script that can zip arbitrary directories and delete an existing output archive if invoked.

Review this skill before installing. The estimation and Excel-generation parts are local and purpose-aligned, but the package includes an unrelated directory-packaging utility that should be removed or clearly separated from the shipped runtime artifact. Users should also expect Chinese-language Excel output and keyword matching unless the publisher adds locale support or documentation.

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 Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 99% confidence
Finding: The declared behavior is limited to test-effort estimation and Excel export, but the detected behavior includes packaging directories into ZIP archives, traversing filesystem paths, deleting an existing output ZIP, and accepting command-line arguments that influence packaging. This mismatch is dangerous because it can hide broader filesystem access and destructive file operations behind an innocuous business-purpose description, increasing the chance of unauthorized data collection or tampering.

Lp3

Medium

Category: MCP Least Privilege
Confidence: 82% confidence
Finding: The skill references reading bundled files and executing a local script, but it does not declare any explicit tool scope such as allowed-tools or permissions. That creates an authorization gap: an agent may read files or invoke capabilities implicitly without the user or platform having a constrained, auditable boundary for what the skill is permitted to access.

Description-Behavior Mismatch

Medium

Confidence: 98% confidence
Finding: The manifest describes a skill for analyzing requirements, breaking down testing tasks, estimating effort, and exporting estimates to Excel. This file performs none of those actions; it packages a directory into a ZIP file and can remove an existing output archive, creating a semantic mismatch between claimed skill behavior and actual code operations.

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The manifest says this skill analyzes requirements, estimates testing effort, and exports results to Excel. This file instead implements a general packaging utility that walks directories, creates ZIP archives, and deletes any existing archive at the destination, which is not justified by the skill's stated business purpose.

Natural-Language Policy Violations

Low

Confidence: 93% confidence
Finding: This Python file contains user-facing natural-language content and keyword logic entirely in Chinese, including feature classification terms and generated rationale text. Because the skill does not provide any language/locale opt-in or explain that it is intended only for a Chinese-language workflow, it can violate the policy against forcing a specific language without user choice.

Natural-Language Policy Violations

Low

Confidence: 98% confidence
Finding: The script prints all status messages in Chinese, which imposes a specific language on users regardless of their locale or preference. This is a natural-language policy concern because there is no opt-in, fallback, or documentation indicating the tool is intentionally limited to Chinese-speaking users.

Static analysis

No suspicious patterns detected.