The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands, including curl for HTTP probing, katana for web crawling, and seq/xargs for race condition testing. It also invokes local Python scripts within the project directory (e.g., brain.py, intel_engine.py) to manage task state and logic.
[EXTERNAL_DOWNLOADS]: The skill performs active network operations against user-specified targets and interacts with external Out-of-Band (OOB) testing services like interactsh and webhook.site to validate vulnerabilities.
[PROMPT_INJECTION]: Instructions use highly authoritative language (e.g., 'MANDATORY', 'VIOLATION = BAN') and 'preambles' to enforce strict behavioral constraints and prompt formatting on sub-agents.
[DATA_EXFILTRATION]: Includes diagnostic payloads for XSS and SSRF that transmit data (such as cookies or metadata) to external servers, which is a standard procedure in security vulnerability confirmation.
[SAFE]: While the skill contains Base64-encoded strings (e.g., YWxlcnQoMSk=) chained with execution patterns (e.g., eval(atob(...))), these are clearly documented as example payloads for remote XSS testing and do not execute within the local environment.

hunt