hunt-file-upload

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). The list includes attacker-controlled COLLAB_HOST callback endpoints and an unknown GitHub PoC repository (mistymntncop/CVE-2023-4863), both high-risk vectors for delivering exploits or hosting malicious payloads.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document is explicitly offensive: it contains actionable, step-by-step techniques to achieve RCE, SSRF (including cloud metadata access), credential and data exfiltration, webshell deployment, and archive abuse — clearly malicious intent and high-risk abuse patterns.

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

• Third-party content exposure detected (medium risk: 0.30). The skill describes a file-upload test workflow targeting endpoints like /upload, /avatar, /profile-picture, /attachment, /import that the agent uses to upload outsider-authored malicious filenames and file contents (e.g., SVG/HTML/DOCX/ZIP payloads), which are then ingested/processed by the application at runtime.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill explicitly documents and instructs numerous security-bypass and exploitation techniques (extension/magic-byte bypasses, .htaccess re-enablement, SSRF to metadata, zip-slip/symlink attacks, creating local payload files and installing tools) which push an agent to perform actions that can compromise system state or bypass protections.