Security audit
Security checks for vulnerabilities and agentic risk
This is a text-rewriting skill whose file access is aligned with editing documents and shows no hidden execution, persistence, or data exfiltration behavior.
Use this skill for intentional rewriting tasks and review diffs before accepting changes, especially for legal, policy, academic, quoted, or compliance-sensitive text where exact wording matters.
The phrase "Or invoke directly when editing documents" does not define specific trigger words, scope, or exclusion conditions for when the skill should activate. In a markdown skill description, this kind of broad activation guidance can cause unintended invocation during ordinary document editing.
The description is broadly scoped to 'editing or reviewing text to make it sound more natural and human-written,' which can cause the skill to activate in many contexts without clear user intent boundaries. In an automated agent environment, that increases the chance of unintended rewriting of sensitive, compliance-relevant, or user-authored text where preserving exact wording matters.
The skill has Write and Edit permissions and is designed to rewrite text, but it does not clearly warn that it may directly modify files rather than just propose edits. That can lead to silent or unexpected changes to repository content, including loss of author intent, corruption of quoted material, or accidental edits to high-integrity documents.
