Security audit

Agent Intelligence Network Scan

Security checks for vulnerabilities and agentic risk

Overview

The skill does not contain executable malware, but it is a sensitive reputation, threat-scoring, and identity-linking tool that encourages trust and investment decisions without enough privacy or accuracy safeguards.

Install only if you are comfortable using a tool that profiles agents across platforms and may query a backend or store reputation data locally. Treat its scores and threat labels as advisory, not as the sole basis for collaboration, investment, moderation, or exclusion decisions.

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

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)

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The trigger text is broad enough to activate on common trust, evaluation, and decision-making queries, which can cause the skill to be invoked in contexts well beyond a narrowly scoped reputation lookup. In this skill's context, overbroad invocation increases the chance that sensitive agent identifiers, collaboration decisions, or risk assessments are routed through this skill unnecessarily, amplifying privacy and decision-manipulation risk.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The documentation describes a backend-connected mode and on-disk caching but does not clearly warn users that queries and agent data may be transmitted to an external service and persisted locally. In a skill centered on reputation and threat intelligence, that omission is more sensitive because queried identities, threat flags, and relationship lookups may themselves be confidential or privacy-impacting.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The documentation exposes threat and reputation scoring capabilities that can be used for profiling, exclusion, or automated trust decisions, yet it provides no warning about sensitivity, false positives, or due-process concerns. In this skill's context, these scores are expressly intended for trustworthiness and collaboration/investment decisions, which makes misuse more likely and amplifies potential harm from inaccurate or biased data.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The API explicitly supports cross-platform identity linking and multi-account correlation, which is sensitive profiling functionality. Documenting and enabling this without any privacy warning, consent requirements, abuse constraints, or guidance on lawful use increases the risk of deanonymization, stalking, or inappropriate surveillance of users across platforms.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The document explicitly states that the system aggregates cross-platform identity, activity, threat intelligence, and manual report data and refreshes it frequently, but it provides no privacy notice, retention limits, consent model, or guidance on handling potentially sensitive profiling data. In a reputation-scoring skill, this omission is security-relevant because operators or downstream users may process personal or quasi-personal data in ways that create privacy, misuse, and compliance risks.

Intent-Code Divergence

Low

Confidence: 89% confidence
Finding: Line L291 states the skill supports 'Moltbook, Moltx, 4claw, Twitter, GitHub', but earlier documentation consistently describes only 4 live platforms and lists data sources limited to Moltbook, Moltx, 4claw, Twitter, identity resolution, and security monitoring. This is a direct documentation contradiction about supported platforms rather than a mere omission.

Static analysis

No suspicious patterns detected.