ctf-sandbox

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a local PowerShell initialization script ../scripts/case-init.ps1. This script appears to be part of the skill's setup procedure to establish security boundaries.
  • [INDIRECT_PROMPT_INJECTION]: As a task orchestrator for CTF and AWD competitions, the skill naturally handles untrusted external data (challenge files, contest prompts).
  • Ingestion points: Processes user input related to CTF, AWD, and competitive security tasks.
  • Boundary markers: The skill explicitly instructs the agent to check auth.status and set specific network profiles (-NetworkProfile lab or offline) to prevent unauthorized internet access.
  • Capability inventory: Capability to trigger initialization scripts and route tasks to over 40 specialized competition sub-skills.
  • Sanitization: The skill relies on environment-level isolation (sandboxing) rather than prompt-level sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 03:53 AM
Security Audit — agent-trust-hub — ctf-sandbox