do-it-gungun - 滚滚判断

• Outcome: Review
• Latest audit: May 29, 2026 · 9:28 PM UTC
• Version: 1.0.0

Overview

This skill is mostly a disclosed life-decision assistant, but it handles sensitive personal decisions with under-disclosed storage, feedback reuse, and one hard-coded external data submission path.

Review before installing or using with real personal details. Avoid entering sensitive financial, relationship, employment, or family information unless you understand where the web/API version sends and stores it. Treat outputs as informal brainstorming, not professional advice, and do not rely on its bundled datasets as verified live data.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection

Findings (42)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 96% confidence
• Finding: The skill is presented as a decision-support markdown spec, but static analysis indicates hidden capabilities for file access, network access, and shell execution without any declared permissions. Undeclared privileged behavior breaks the trust boundary for users and can enable local data exposure, remote data exfiltration, or arbitrary command execution if the skill is invoked in an agent runtime.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 98% confidence
• Finding: There is a strong mismatch between the stated purpose ('life-choice judgment assistance') and the detected operational behavior, which includes serving HTTP endpoints, reading local files, scraping or generating datasets, and running external scripts. This kind of hidden functionality is dangerous because users may provide sensitive personal information believing they are using a simple advisory skill, while the skill may actually collect, process, expose, or transmit data far beyond that expectation.

Intent-Code Divergence

Medium

• Confidence: 96% confidence
• Finding: The module and function docstrings state that the script collects real decision cases from internet sources, but the implementation only returns hardcoded sample records and saves them as if they were crawled data. This is a data integrity and provenance issue: downstream components or users may trust fabricated content as real-world evidence, leading to misleading analysis, bad decisions, or false claims about dataset origin.

Intent-Code Divergence

Low

• Confidence: 87% confidence
• Finding: The emitted records are marked `pending_analysis`, which implies they are authentic collected cases awaiting downstream review, but they are fabricated placeholders. In this skill context, that mismatch increases the chance that later AI analysis treats synthetic examples as real user-generated cases, degrading trust and potentially contaminating any decision-support outputs.

Intent-Code Divergence

Medium

• Confidence: 86% confidence
• Finding: The function claims to retrieve live public data but actually returns hard-coded sample values, which can mislead downstream components or users into treating fabricated data as current external intelligence. In a life-decision skill, this integrity failure is more dangerous because users may rely on the output for relocation or career choices under false assumptions of freshness and provenance.

Intent-Code Divergence

Medium

• Confidence: 86% confidence
• Finding: The housing crawler presents itself as sourcing current market prices from Anjuke while returning built-in values, creating a data provenance and trust violation. Because housing costs strongly influence life and relocation decisions, consumers of this skill may make harmful financial decisions based on stale or fictional values.

Intent-Code Divergence

Medium

• Confidence: 87% confidence
• Finding: The GDP function claims to obtain statistics from official or government sources but instead emits static estimated values, which undermines the integrity of socioeconomic scoring derived from it. In this skill context, that misinformation can propagate into city rankings and recommendations that appear authoritative, increasing the likelihood of user harm through bad decisions.

Description-Behavior Mismatch

High

• Confidence: 92% confidence
• Finding: The file’s functionality materially diverges from the skill’s declared purpose: instead of only supporting AI judgment for life choices, it implements salary data collection, local dataset generation, and manual data entry. This kind of scope mismatch is dangerous because it hides materially different capabilities from reviewers and users, making it easier for undeclared data collection or later expansion into broader scraping behavior to evade scrutiny.

Context-Inappropriate Capability

Medium

• Confidence: 88% confidence
• Finding: This file includes third-party job-site crawling logic that is not justified by the stated skill purpose, expanding the skill’s real capabilities beyond what users and reviewers would expect. Even though the current functions are mostly placeholders, they establish a framework for unauthorized scraping of external services, which can create compliance, trust, and data-handling risks once completed.

Context-Inappropriate Capability

Medium

• Confidence: 84% confidence
• Finding: Persistent local dataset creation and modification broaden the skill from advisory behavior into data collection and storage, which is outside the stated scope. This is risky because undeclared storage can accumulate sensitive or regulated data over time and complicates review of what the skill actually retains or modifies on disk.

Description-Behavior Mismatch

Medium

• Confidence: 92% confidence
• Finding: The page collects sensitive personal decision-making details, including relationship, career, family, age, housing, and financial context, then transmits them to an external server. This exceeds what a user may reasonably expect from a local-looking test page and creates privacy and data-governance risk, especially because the transmission target is a hard-coded third-party IP address.

Context-Inappropriate Capability

Low

• Confidence: 83% confidence
• Finding: The page loads Tailwind from an external CDN at runtime, which introduces a third-party supply-chain and privacy dependency. If the CDN response is tampered with, blocked, or used for tracking, the page could execute untrusted script in the origin of the app and expose locally stored judgment history.

Intent-Code Divergence

Medium

• Confidence: 96% confidence
• Finding: The UI explicitly suggests that the system is collecting data, comparing options, and generating a judgment, but the fallback path returns a canned recommendation unrelated to the user's actual situation. In a decision-making skill handling career, relationship, and investment questions, this can mislead users into acting on fabricated analysis while believing it is evidence-based.

Missing User Warnings

Medium

• Confidence: 87% confidence
• Finding: The document encourages deployment via Vercel and separately describes plans for storing user history and integrating a real AI API, but it provides no privacy, access-control, or data-handling warnings. In this skill's context, users are expected to submit personal life-decision details, so deploying a web app without clear safeguards can lead to unintended public exposure, overcollection, or third-party disclosure of sensitive user data.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The document explicitly defines an archival/knowledge-base step for user decision outputs but does not warn users that their questions, decisions, or outcomes may be retained and reused. In a skill centered on personal life choices, this can expose sensitive personal, career, relationship, or financial information to unintended persistence and downstream reuse.

Missing User Warnings

Medium

• Confidence: 98% confidence
• Finding: The continuous learning workflow says user feedback and results are used for review and knowledge-base updates, but it gives no consent, retention, or privacy warning. Because users may provide highly sensitive follow-up outcomes, this creates a real risk of undisclosed secondary use of personal data.

Missing User Warnings

Medium

• Confidence: 98% confidence
• Finding: The post-decision process directly instructs reuse of user feedback and results in an ongoing optimization loop without any privacy or data-use disclosure. In context, the skill is designed for major personal decisions, so retained feedback may contain intimate and high-sensitivity information that users would not reasonably expect to be persisted.

Vague Triggers

Medium

• Confidence: 97% confidence
• Finding: The skill is framed as applicable to essentially any difficult life decision, including careers, relationships, investments, and housing, without gating, exclusions, or escalation rules for high-stakes situations. This is dangerous because it encourages overreliance on a generic decision agent in domains where poor advice can cause significant financial, emotional, or life-impacting harm.

Missing User Warnings

Medium

• Confidence: 98% confidence
• Finding: The description promotes direct guidance on consequential choices such as investments and major life decisions but does not prominently warn users about the limits of the advice at the point of use. Even though a brief disclaimer appears later in the output template, the overall framing and branding ('you just do it') can push users toward taking action without adequate caution.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The product design explicitly asks users to submit highly sensitive personal context such as career status, concerns, preferences, and other free-form background information, and later references data saving and personal history, but provides no visible privacy notice, retention limits, consent flow, or handling constraints. In a decision-advice product, users are likely to disclose intimate financial, employment, relationship, or life-choice details, so omission of user-facing privacy safeguards increases the risk of overcollection, unexpected storage, and misuse or exposure of personal data.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The design includes a case library and sharing functionality displaying realistic user scenarios with identifiable contextual details, but does not warn users that their submissions may become visible to others or be shared externally. Because these scenarios involve sensitive life decisions and may contain age, city, salary, career history, and personal circumstances, publication without explicit consent or strong anonymization can expose users to privacy harm, re-identification, and secondary misuse.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The plan explicitly includes collecting user problem descriptions, background information, execution feedback, result records, and satisfaction scores, but it provides no privacy notice, retention limits, consent model, or safeguards for potentially sensitive personal data. Because the product targets life decisions such as career, relationships, investment, and personal life, these records may contain highly sensitive information whose mishandling could expose users to privacy harm, profiling, or unauthorized reuse.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The document proposes AI dialogue, personalized recommendations, community features, and a user-feedback loop, all of which imply behavioral profiling and analysis of user-submitted data, yet it does not disclose those implications or establish boundaries for automated decision support. In this skill context, the recommendations concern consequential personal choices, so undisclosed profiling can materially affect users while also increasing privacy and fairness risk.

Missing User Warnings

Medium

• Confidence: 83% confidence
• Finding: The plan includes external API usage, a data API business model, automated data collection/update mechanisms, and third-party AI integration, but it does not address external data transmission, provenance validation, integrity checks, or downstream data-sharing risks. That omission is dangerous because the system may send user queries or sensitive decision context to outside services and may ingest untrusted data that can affect recommendations or expose users to unauthorized disclosure.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The project summary explicitly mentions saving user history, but provides no notice about what data is retained, for how long, or how users can control deletion. In a decision-support product where users may submit sensitive personal, career, financial, or relationship information, undocumented persistence increases privacy and compliance risk and can lead to unintended exposure of user data.

VirusTotal

67/67 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.