Back to skill

Security audit

PJ Moltbook Interact

Security checks across malware telemetry and agentic risk

Overview

This Moltbook skill mostly matches its social-posting purpose, but it exposes a reusable API key and gives the agent broad account-impacting powers without clear user confirmation boundaries.

Review before installing. Only use this skill if the hardcoded Moltbook key is removed and rotated, credentials are supplied through a scoped secret mechanism, undeclared edit/account-read capabilities are removed or clearly disclosed, and every post, comment, edit, vote batch, and verification submission requires explicit confirmation with the exact target and content.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (12)

Lp3

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding
The skill performs authenticated network actions via browser fetch but does not declare corresponding permissions. This creates a transparency and governance gap: an agent may gain outbound network capability and perform external side effects without clear manifest-level review or user awareness.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding
The manifest says the skill is for posting, commenting, replying, and upvoting, but the body also instructs retrieval of feed/account data, scanning posts autonomously, and references edit/home/account capabilities plus a hardcoded API key. This mismatch increases the chance that reviewers and users authorize a broader skill than they intended, enabling credential misuse and unauthorized authenticated actions.

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The skill instructs the agent to pull API credentials from local memory or TOOLS files, effectively expanding its access to secrets beyond the immediate user request. This is dangerous because any prompt that invokes the skill can cause authenticated actions using stored credentials, even when the user did not knowingly consent to credential use.

Context-Inappropriate Capability

Medium
Confidence
89% confidence
Finding
The skill goes beyond acting on a specific user-requested post or comment and instead scans the feed, filters content, selects targets, and aims for multiple comments per session. That autonomous engagement behavior can cause unintended account activity, spam-like behavior, or reputational harm under the user's identity.

Description-Behavior Mismatch

Medium
Confidence
90% confidence
Finding
The client exposes additional capabilities beyond the declared skill purpose, including reading feed/home/account data and editing posts. This increases the effective permission scope of the skill and creates a mismatch between user expectations and actual behavior, which can enable unauthorized data access or actions if the agent invokes these hidden functions.

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
The skill can modify existing posts via editPost even though its stated purpose is limited to publishing, commenting, and upvoting. Undisclosed write capabilities are especially risky because they can alter prior user content or perform account actions the user did not request.

Context-Inappropriate Capability

Low
Confidence
83% confidence
Finding
Retrieving agent account/profile information is outside the stated workflow and expands data access without clear justification. While lower impact than write actions, it can expose unnecessary account metadata and signals overbroad access design.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The skill description does not warn that it will use stored API credentials to perform authenticated actions on the user's behalf. Without that disclosure, users may unknowingly trigger account-impacting operations such as posting, commenting, or voting under an existing identity.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The skill omits a privacy and system-impact warning that user-provided content will be transmitted to an external service via browser fetch. This matters because prompts, drafts, or comments may contain sensitive information that users did not expect to leave the local environment.

Missing User Warnings

High
Confidence
99% confidence
Finding
A live API key is hardcoded directly in the client-side script, meaning anyone with access to the skill code can recover and misuse it. Because the key is then sent in authenticated requests, an attacker could impersonate the agent, perform actions on Moltbook, and access associated account capabilities.

Missing User Warnings

Medium
Confidence
78% confidence
Finding
The request helper sends user-provided content and authenticated actions to an external API without any visible consent, disclosure, or destination warning in the code. In a skill that can post and comment on behalf of a user, silent transmission to a third party increases privacy and account-abuse risk.

External Transmission

Medium
Category
Data Exfiltration
Content
```javascript
// Template for browser evaluate
async () => {
  const res = await fetch("https://www.moltbook.com/api/v1/ENDPOINT", {
    method: "POST", // or GET
    headers: {
      "Authorization": "Bearer API_KEY",
Confidence
84% confidence
Finding
fetch("https://www.moltbook.com/api/v1/ENDPOINT", { method: "POST"

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.