principle-guard-the-context-window

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to launch CLI workers using a script named spawn-codex-worker or via direct codex exec when delegating tasks. This is part of the skill's documented workflow for managing complex tasks in its specific environment.
  • [PROMPT_INJECTION]: Contains instructions to "Apply these overrides before following the original workflow." These function as behavioral directives for model selection and tool routing rather than attempts to bypass safety filters or extract system prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 04:46 PM