Security audit
Security checks for vulnerabilities and agentic risk
This is a note-editing helper that only provides instructions for rewriting or answering about selected text, with no code execution, persistence, credential access, or hidden data flow.
Install this only in note-editing flows where edits are clearly scoped to a user selection and can be previewed or undone. Be careful with ambiguous prompts because the skill may return replacement text instead of an explanatory answer.
The skill is described for broad use in note editors and AI writing assistants without defining clear activation boundaries, increasing the chance it is invoked in contexts where silent text modification is inappropriate or unexpected. In an editor-integrated assistant, ambiguous triggering can cause unintended content changes or user confusion, especially when handling selected text automatically.
The top-level description markets seamless integration and silent replacement behavior but does not prominently warn that instruction mode may directly replace selected text without explanatory output. In a note-editing context, this increases the risk of users unintentionally invoking destructive edits and losing or altering content they expected only to discuss.
The skill says it 'automatically distinguishes two modes' but provides only high-level examples rather than deterministic rules, so ordinary requests may be misclassified as edit commands or questions. Because instruction mode outputs replacement text only, a misclassification can silently overwrite selected content without explanation or confirmation.
No suspicious patterns detected.
