ClawHub
Back to skill

Security audit

adaptador-formatos-f100k

Security checks for vulnerabilities and agentic risk

Overview

This skill analyzes a user-provided social video link to adapt its content format, with disclosed use of social scraping tools and limited local note-saving.

Before installing, understand that this skill may send user-provided social media URLs to scraping or public-page tools, read captions/thumbnails/comments, and save the resulting adapted ideas plus the source link in local idea files. Avoid using it on private or sensitive posts unless you are comfortable sharing screenshots or context manually.

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 (3)

Vague Triggers

Medium
Confidence
95% confidence
Finding
En un archivo markdown/manifest, los disparadores deben ser específicos para evitar invocaciones accidentales. Aquí se listan ejemplos concretos, pero la cláusula abierta "o cualquier variación" amplía el alcance de forma ambigua y sin ejemplos negativos o límites claros.

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The OpenClaw section says Paso 1 should be done without a logged-in browser and specifically use Apify or public-page tooling, while the later process instructs opening the link 'logueada' with agent-browser for Instagram/TikTok. These instructions point to materially different behaviors for the same step, creating intent ambiguity about whether authenticated browser access is part of the skill.

Missing User Warnings

Medium
Confidence
87% confidence
Finding
The skill instructs the agent to access third-party content, inspect comments/captures, and persist outputs to local files, but it does not clearly require user-facing disclosure or consent about data handling and retention. In practice, this can lead to collection and storage of external content metadata, user-provided screenshots, and derived notes without transparent notice, creating privacy and governance risk—especially when posts are private or screenshots may contain personal data.

Static analysis

No suspicious patterns detected.