🛡️

Tools Overview

The agent has direct access to all tools. No sub-agent spawning - simplicity and effectiveness.

Direct Access Philosophy

Unlike multi-agent systems, the main agent has direct access to every tool. This eliminates coordination overhead and allows for faster, more coherent decision-making. The agent's behavior is controlled by prompts, not code - same engine, different expertise.

💻

System Tools

Core system interaction

run_cmdread_filewrite_file
  • •Execute shell commands
  • •Read configuration files
  • •Create scripts and payloads
🔍

Reconnaissance

Information gathering and discovery

parse_nmapsearch_cveweb_search
  • •Parse scan results
  • •Find known vulnerabilities
  • •Research exploit techniques
🎯

Exploitation

Active attack execution

browse_urlfill_submit_formhash_crack
  • •Analyze web pages
  • •Test form inputs
  • •Crack password hashes
🌐

Network Attack

Network-level operations

arp_spoofmitm_proxypacket_sniff
  • •ARP spoofing
  • •Man-in-the-middle
  • •Traffic capture
📊

State Management

Tracking and reporting

add_findingadd_targetget_state
  • •Record vulnerabilities
  • •Track discovered hosts
  • •View current progress
⚙️

Resource Control

Background process management

bg_processbg_readbg_stop
  • •Start listeners
  • •Monitor output
  • •Cleanup processes

Approval Flow

Tool Call
→
Scope Check
→
Approval Gate
→
Execute
→
Log

Self-Recovery Protocol

When tools fail, the agent analyzes the error and attempts recovery:

  • → missing parameter → Check schema, add missing values
  • → command not found → Try alternative tool or install
  • → connection refused → Verify port/protocol, try alternatives
  • → permission denied → Try sudo or different approach
  • → unknown error → Search web for solution, adapt