web-file-upload

Warn

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONDATA_EXFILTRATIONOBFUSCATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions and specific payloads to achieve arbitrary command execution on target systems. This includes the use of PHP system calls (e.g., system($_GET[0])) and exploiting vulnerabilities in image processing libraries such as ImageMagick (ImageTragick) and Ghostscript to gain RCE.
  • [PRIVILEGE_ESCALATION]: Detailed guidance is provided for using path traversal sequences (e.g., ../../var/www/html/shell.php) to escape restricted upload directories. This technique is intended to overwrite files in the webroot or other sensitive locations, effectively escalating the attacker's influence over the target server.
  • [DATA_EXFILTRATION]: The skill describes methods for Server-Side Request Forgery (SSRF) and Out-of-Band (OOB) callbacks. These are achieved by uploading malicious SVG or PDF files that trigger the server to make external requests or leak information to attacker-controlled infrastructure.
  • [OBFUSCATION]: The instructions leverage various encoding and system-specific tricks to bypass security filters. This includes URL-encoded null bytes (%00) and spaces (%20), NTFS alternate data streams (::$DATA), and hex-encoded magic bytes (e.g., \xFF\xD8\xFF\xE0) used to masquerade malicious scripts as legitimate image files.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted file data (e.g., resumes, avatars, CSV/XML imports) as part of its primary function.
  • Ingestion points: Documented in SKILL.md as any file upload vector (avatars, attachments, profile images, document processors).
  • Boundary markers: None identified; the skill does not suggest using delimiters to separate untrusted file content from instructions.
  • Capability inventory: The skill uses tools like burp and exiftool to manipulate files and intercept traffic, and it provides instructions for shell execution and network-based callbacks.
  • Sanitization: The skill focus is entirely on bypassing existing server-side sanitization and validation logic rather than implementing it.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 20, 2026, 11:02 PM
Security Audit — agent-trust-hub — web-file-upload