A bunch of skills on playbooks got flagged as unsafe because the SKILL.⁠md instructs the agent to download and run unsigned executable files. Hoping other directories/registries add prompt injection checks soon - feels like a disaster waiting to happen...