Amazing Idea Generator

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:45 PM UTC
• Version:
  0.6.0

Overview

This is a disclosed idea-generator skill with local memory and optional web search, not a hidden or destructive tool.

Install only if you are comfortable with a creative tool that remembers local interactions, stores submitted ideas and logs, reads OpenClaw profile/memory files for personalization, and may use DuckDuckGo for trend searches. Do not submit confidential ideas, medical/legal/financial details, private meeting content, or sensitive profile data unless you are prepared for it to be stored locally and possibly used in later outputs.

SkillSpector

By NVIDIA

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 Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (32)

Context-Inappropriate Capability

Medium

• Confidence: 95% confidence
• Finding: The skill reads broad user context from USER.md and recent memory files to personalize idea generation, even though that access is not necessary for a basic idea generator. This creates an unnecessary privacy boundary violation and can expose sensitive personal data to prompts, outputs, logs, or downstream model calls without clear user consent.

Context-Inappropriate Capability

Medium

• Confidence: 91% confidence
• Finding: The skill adds outbound network access via DuckDuckGo searches for trending ideas based on user input, expanding the attack surface beyond local idea generation. User-supplied text can be transmitted to an external service without a clear consent step, creating privacy and data-handling risk.

Context-Inappropriate Capability

Medium

• Confidence: 92% confidence
• Finding: The deep analysis feature performs additional web searches using idea names and analysis-oriented queries, turning a simple generator into an external reconnaissance workflow. This can leak user interests or custom content to third parties and broadens data exposure without necessity or transparency.

Intent-Code Divergence

Medium

• Confidence: 95% confidence
• Finding: The code advertises a 5-agent voting system but never performs real agent evaluation; it uses deterministic pseudo-random scores seeded from the idea name. This can mislead users into trusting fabricated review results as if they came from independent evaluators, which undermines integrity and can bias downstream decisions based on false provenance.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill explicitly documents persistent storage of user-related data but does not clearly warn users that personalization relies on reading USER.md and memory files. This creates a privacy risk because users may unknowingly expose preferences, history, or other sensitive context that can be retained and reused across sessions.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The documentation says user-submitted ideas are stored, but it does not clearly notify users at submission time that their content will be persisted. Users may include personal, proprietary, or sensitive information in submissions, leading to unintended retention and later disclosure.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The skill advertises command logging without a clear warning that user interactions will be recorded. Logging prompts and commands can capture sensitive content, behavioral patterns, or identifiers, especially in a conversational tool that encourages free-form input.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The optional DuckDuckGo integration can transmit user prompts or inferred intent over the network, yet the documentation does not clearly warn users about this external disclosure. Even derived search queries can reveal interests, plans, or confidential topics to third-party services.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The legal-advice skill advertises 'basic legal consultation and advice' plus contract review, but provides no disclaimer that outputs are not a substitute for licensed legal counsel and may be incomplete or jurisdiction-specific. In a high-risk domain, this can cause users to rely on inaccurate guidance for contracts, disputes, or compliance decisions.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The health-management skill promises personalized advice based on health data without warning about sensitive data handling or the non-professional nature of the guidance. Users may disclose medical information and act on unsafe recommendations without consulting clinicians, creating privacy and safety risks.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The investment-advice skill offers investment recommendations and risk assessment without disclosing financial risk, uncertainty, or that outputs are not professional financial advice. This can mislead users into making consequential financial decisions based on unreliable or unsuitable recommendations.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: This section describes a mental health diary that records emotional changes and generates suggestions, which involves sensitive mental-health data. The file provides no privacy notice, data minimization guidance, consent language, or handling constraints, creating risk of over-collection and unsafe downstream use of highly sensitive personal information.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: These ideas cover medication reminders and health check report management, both of which involve sensitive medical records and potentially regulated health information. The content normalizes collection, OCR extraction, and analysis of this data without any mention of confidentiality, access control, retention, user consent, or breach risks.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: Multiple entries propose LLM-generated health analysis, workout plans, injury-prevention advice, and health Q&A, but they do not warn that outputs may be inaccurate and are not professional medical advice. In a health context, users may rely on generated recommendations for exercise, medication-adjacent decisions, or symptom interpretation, increasing the chance of physical harm or delayed professional care.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The entry describes child location monitoring, geofencing, and emergency alerts without any visible privacy safeguards, consent requirements, retention limits, or abuse-prevention language. Because the context is parenting and involves monitoring minors, omission of safety controls increases the risk of privacy invasion, unauthorized tracking, or misuse in coercive/family-surveillance scenarios.

Missing User Warnings

Medium

• Confidence: 85% confidence
• Finding: The file describes capabilities such as message monitoring, contact syncing, location-based recommendations, business card OCR, social media management, and account backup, but provides no mention of consent, minimization, retention, or privacy safeguards. In a social-data context, this omission can normalize collection and processing of sensitive personal data without clear user awareness, increasing the risk of privacy violations or misuse when these ideas are implemented.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The WeChat voice-processing tool explicitly describes intercepting voice messages, converting them via ASR/TTS, mixing them with music, and sending the transformed audio back into group chats, but it does not include clear user consent, privacy, or disclosure safeguards. This is dangerous because voice messages may contain personal or sensitive content, and automatic retransmission to group chats can expose or misuse that content in ways users did not authorize.

Missing User Warnings

Medium

• Confidence: 86% confidence
• Finding: The meeting meme generator processes meeting recordings through transcription and extracts notable phrases for meme generation, but the description omits a strong warning that meeting audio and transcripts can contain confidential business, personnel, or regulated information. This creates risk of unauthorized disclosure, embarrassment, or policy violations if sensitive content is transformed and shared as memes.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The WeChat voice tool describes downloading voice messages, transcribing them, and reposting transformed audio into group chats, but does not mention consent, notice, or privacy safeguards. This creates a real risk of processing private communications and redistributing derived content without participants' knowledge, especially in group settings where messages may contain sensitive or identifying information.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The recipe screenshot tool performs OCR on uploaded images and sends extracted text to the OpenAI API without an upfront disclosure that content leaves the local environment. Screenshots can easily contain unrelated personal or sensitive information, so silent external transmission can expose user data beyond what they reasonably expect.

Missing User Warnings

Low

• Confidence: 88% confidence
• Finding: The package tracking tool sends tracking numbers to an external logistics API without clearly informing the user. Tracking numbers are lower sensitivity than messages or meeting audio, but they can still reveal purchase activity, delivery patterns, and location-related information if mishandled.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The meeting meme generator processes recordings and transcripts from meetings without any explicit warning about sensitive business, legal, HR, or personal content. In this context, the absence of consent and confidentiality controls is particularly dangerous because meetings often contain proprietary information, credentials, strategic decisions, or regulated data.

Missing User Warnings

High

• Confidence: 92% confidence
• Finding: This item explicitly describes activity tracking and monitoring to generate work logs, which can enable collection of sensitive employee behavioral data without transparency or consent safeguards. In a workplace context, such monitoring raises significant privacy and insider-risk concerns because it may capture detailed user activity, performance signals, and potentially confidential content.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: Sensitive user profile and memory files are read for personalization with no user-facing disclosure or informed consent. In a skill whose primary purpose is idea generation, hidden access to personal files is contextually more dangerous because users would not reasonably expect it.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: Web search sends user-provided content to an external search provider without warning or approval. Even seemingly harmless requests can contain sensitive interests, project names, or private text, making silent transmission inappropriate for this skill context.

VirusTotal

65/65 vendors flagged this skill as clean.