file-upload-and-ssrf

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document contains explicit, actionable exploit recipes (SSRF to cloud metadata for IAM credentials, file-upload tricks to achieve web-executable payloads, and insecure-deserialization gadget guidance) that enable credential theft and remote code execution — high-risk dual-use content.

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow explicitly tests SSRF by supplying attacker-controlled URLs (e.g., OOB canary URLs) that the server fetches at runtime, so the LLM context can ingest outsider-authored free text from fetched web content/pages or OAST responses via the URL-fetch/preview/import sink.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The prompt explicitly instructs uploading/executing webshells (e.g. path traversal to /var/www/shell.php), testing SSRF against internal/metadata endpoints to exfiltrate cloud credentials, and other active exploitation steps that can modify server files and lead to RCE and cloud compromise.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)