Plenty of Claws

• Outcome: Pass
• Latest audit: Jul 10, 2026 · 3:21 AM UTC
• Version: 1.0.0

Overview

This skill is a simple local AI-agent profile directory, but users should treat its profile storage as experimental and avoid sensitive profile data.

Install only if you are comfortable with agent profile information being saved in a local JSON file and potentially shown through the skill's browse/search commands. Do not put sensitive personal details in profiles. This version appears buggy: saved profiles may not load correctly, new sign-ups may overwrite prior profile data, and manual test scripts can delete a local profiles.json file.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• 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 (9)

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 91% confidence
• Finding: The skill documentation promises dating-profile creation, browsing, and compatible-match searching, but the described implementation also performs persistent local file I/O and appears not to load existing profiles correctly. This mismatch can mislead users and host systems about data handling and capability boundaries, which is dangerous because undeclared persistence affects privacy expectations and broken behavior can cause unsafe assumptions about what data is being accessed or retained.

Intent-Code Divergence

Medium

• Confidence: 98% confidence
• Finding: The skill defines persistence helpers but never calls loadProfiles(); instead it initializes profiles as an empty array and then saves that array after signup. This causes existing stored profiles to be ignored and overwritten, leading to silent data loss and making profile enumeration/view features unreliable.

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The documented trigger phrases are very generic natural-language commands such as "Sign up" and "View profile," which are likely to appear in ordinary conversation and may cause unintended skill activation. In an agent environment, accidental invocation can lead to profile creation, profile disclosure, or other state-changing actions without clear user intent.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The README advertises persistent storage in `profiles.json` but does not clearly warn users that profile data is written to a local file and retained across sessions. This can create privacy and data-handling risks because users may provide personal or sensitive profile content without understanding where it is stored or who else on the system can access it.

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: Using a generic trigger like "Help" increases the chance of accidental invocation during ordinary conversation, which can cause the skill to activate outside user intent. In a skill that exposes or persists profile data, unintended activation may reveal profile listings or create confusion about whether user input is being treated as commands.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The skill mentions persistent storage in features, but it does not clearly warn users in the main description and getting-started flow that profile data is retained in a local file across sessions. For a dating/social-profile context, omission of clear storage disclosure increases privacy risk because agents or operators may share personal or identifying profile content without understanding retention behavior.

Missing User Warnings

Medium

• Confidence: 81% confidence
• Finding: The skill stores user/agent profile information on disk without disclosing that persistence behavior to users. In a dating-style profile system, names, bios, interests, and identifiers may be sensitive, so undisclosed storage creates privacy and consent risks even if no direct exfiltration is shown.

Missing User Warnings

Medium

• Confidence: 98% confidence
• Finding: This test script unconditionally deletes profiles.json at startup if it exists, which can destroy persisted profile data without confirmation, backup, or isolation to a test-only path. In the context of a social-profile skill, that means real user/agent data could be erased if the test is run in the same directory/environment as production data.

Missing User Warnings

Medium

• Confidence: 98% confidence
• Finding: The cleanup step deletes profiles.json again at the end of execution without verifying that the file is test-generated or safe to remove. If this script runs in a shared or production-like environment, it can cause silent loss of all profile records and interfere with concurrent use of the skill.

VirusTotal

35/35 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.