Security audit

analizador-referencias-virales-f100k

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed social-video research workflow that uses external scraping/transcription services and saves reports, with no evidence of hidden, destructive, or deceptive behavior.

Before installing, make sure you are comfortable with public social-media URLs, handles, metrics, and transcript-derived analysis being sent through Apify/Supadata, posted back in Telegram, and saved locally under informes/. For sensitive research, ask the agent not to save or share externally unless needed.

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 (5)

Description-Behavior Mismatch

Medium

Confidence: 89% confidence
Finding: The skill documentation says the summarized table goes to Telegram in addition to being saved under `informes/`, which expands the data flow beyond the main output locations a user might expect. Sending scraped URLs, handles, transcripts, or analysis to a messaging channel increases the chance of unanticipated disclosure, especially if the Telegram chat includes other participants or is retained externally.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The activation phrases are broad enough to match ordinary research or ideation requests, which can cause the skill to invoke scraping, transcription, and persistent reporting when the user may have intended a simple answer. That ambiguity matters because this skill performs external API calls, processes third-party content, and stores outputs, so accidental activation can trigger unnecessary data collection and side effects.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill mandates always writing a report to `informes/` even if the user did not request persistence, creating an undisclosed storage side effect. Persistent files can retain scraped third-party content, usernames, URLs, and derived analysis longer than the user expects, increasing privacy, retention, and cross-session exposure risk.

Intent-Code Divergence

Low

Confidence: 95% confidence
Finding: Line L016 instructs that keywords must be run one after another, not in parallel. Later, the skill documentation says to launch one run per keyword in parallel, creating an active contradiction in operator intent and expected behavior.

Intent-Code Divergence

Low

Confidence: 95% confidence
Finding: Line L131 says to launch one run per keyword in parallel, but earlier guidance at L016 requires processing keywords one after another. This is a direct contradiction in the documented operating procedure.

Static analysis

No suspicious patterns detected.