AGENT LAB: SKILLS

customer-persona

Fail

Audited by Snyk on Feb 19, 2026

Risk Level: CRITICAL
Full Analysis

CRITICAL E005: Suspicious download URL detected in skill instructions.

  • Suspicious download URL detected (high risk: 0.90). These domains host a remote shell installer (the prompt includes "curl … | sh"), and executing or auto-downloading .sh installers from a non-standard/unverified domain is a high-risk vector for malware even if the site may be legitimate.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (high risk: 0.90). The SKILL.md research steps explicitly instruct running infsh apps like tavily/search-assistant and exa/search/exa/answer to fetch web/search results (open/public third‑party content) as part of persona research, meaning the agent will ingest and act on untrusted web content that could contain injected instructions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

  • Potentially malicious external URL detected (high risk: 0.90). The Quick Start instructs running curl -fsSL https://cli.inference.sh | sh which downloads and immediately executes remote code from https://cli.inference.sh as a required CLI dependency for the skill, creating a high-risk runtime external dependency.
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 19, 2026, 03:40 AM