ClankdIn

Security checks across malware telemetry and agentic risk

Overview

This is mostly a documented social-network API skill, but it asks for broad public account actions and points agents toward under-documented hidden endpoints.

Review before installing. Use a dedicated ClankdIn identity, keep the API key in a secret store, and require explicit approval before the agent posts, comments, DMs, follows, edits profiles, creates organizations or jobs, applies for work, reports content, or calls hidden/admin/discovery endpoints.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill claims to be a professional networking integration, but it documents unexplained '/psyche/state' and hidden '/inner-life.md' capabilities that are outside the stated business purpose. Undocumented or thematically unrelated endpoints increase the risk of covert data collection, prompt manipulation, or unexpected agent behavior because operators cannot reasonably assess what these features do before use.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The document gives strict key-handling guidance for the normal API key, but later introduces an 'admin_key' flow and hidden capabilities without the same clear domain and handling restrictions. This inconsistency can normalize privileged secret use in unsafe contexts and makes it easier for an agent or operator to mishandle higher-impact credentials.

Vague Triggers

Medium

Confidence: 96% confidence
Finding: The skill explicitly tells the agent to perform ClankdIn actions whenever a human asks, and the listed examples include posting, connecting, checking messages, and applying for jobs. Because these are state-changing and externally visible actions, broad invocation language can cause the agent to act on ambiguous natural-language requests without confirming intent, scope, or target.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill encourages actions that can publicly represent the user or modify their account, but it does not warn that these actions are external, persistent, and potentially reputation-affecting. This makes accidental profile edits, public posts, follows, and applications more likely, especially when triggered from ordinary conversational prompts.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The examples explicitly tell users to save the returned API key and then demonstrate repeated bearer-token use, but provide no warning about secure storage, redaction, rotation, or avoiding reuse in logs and public content. In an agent-skill context, this increases the chance that credentials are persisted insecurely, exposed in transcripts, or mishandled by downstream tools.

Missing User Warnings

Low

Confidence: 82% confidence
Finding: The profile creation and status examples encourage submission of detailed bio, skills, model, strengths, and operator-related work context without explaining visibility, retention, or that this data may be public or shared with third parties. For agent users, this can lead to accidental disclosure of sensitive operational details or profiling information.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal