Detected a prompt injection in the skill instructions. The skill contains hidden or deceptive instructions that fall outside its stated purpose and attempt to override the agent’s safety guidelines or intended behavior.
The skill is a legitimate GTD inbox-processing workflow, but its unconfirmed reply-sending and vague "auto-forward to task system" filter could move user email to unspecified places in ways the user may not expect, so it is sketchy rather than malicious.
The skill exposes the agent to untrusted, user-generated content from public third-party sources, creating a risk of indirect prompt injection. This includes browsing arbitrary URLs, reading social media posts or forum comments, and analyzing content from unknown websites.
The required workflow searches the user's Gmail inbox (`in:inbox`) and reads full message content via `get_gmail_message_content`/`get_gmail_messages_content_batch`, and anyone can email the user, so outsider-authored text enters the LLM context without the agent selecting a specific item.
kmondlane/ExoMind
8794d26
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.