Security audit
Security checks for vulnerabilities and agentic risk
This markdown-only skill is purpose-aligned for disciplined follow-through, but users should know it encourages quieter, more autonomous work once a task direction is approved.
Install this only if you want the assistant to be more autonomous after you have approved a direction. Keep an eye on tasks involving commits, durable rules, or skill/project-document changes, and ask for explicit checkpoints when the work is sensitive or high-impact.
The invocation criteria are broad enough to trigger on ordinary user corrections such as requests to keep moving, prioritize, or reduce chatter. That can cause the assistant to enter a more forceful execution mode outside clearly bounded contexts, increasing the chance of unintended actions or reduced confirmation before edits and commands.
The skill explicitly encourages silent execution with short, infrequent updates while continuing to edit files or take concrete actions. In an agent setting, this reduces user visibility into impactful operations and can normalize behavior where changes occur without clear contemporaneous notice, making misuse or overreach harder to detect.
No suspicious patterns detected.
