Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent book-writing assistant that saves manuscript organization data locally, with no hidden code, networking, credential use, or destructive behavior found.
Install this only if you want the assistant to maintain a local book-data.json record of manuscript details, research notes, and continuity data. Treat that file as potentially sensitive unpublished writing material and review or delete it according to your privacy needs.
The trigger instructions are intentionally very broad, including casual phrases like 'I'm working on my book' and many generic writing-related terms. This can cause the skill to activate in contexts where the user did not intend persistent manuscript tracking or creative-assistant behavior, increasing the chance of unnecessary data collection, context hijacking, or user confusion.
The skill directs persistent storage of detailed manuscript data in book-data.json as the 'single source of truth' but does not clearly warn users up front that their content will be saved locally. Because book manuscripts may contain sensitive personal material, unpublished intellectual property, or research notes, silent persistence creates a meaningful privacy and confidentiality risk.
The instruction 'Never use em dashes' imposes a fixed style constraint on responses regardless of user preference. This is a natural-language policy concern because it forces a specific language/locale-related writing convention without offering the user a choice, except for drafted manuscript passages.
The eval explicitly expects the skill to write to a local file (book-data.json) from a natural-language request, but it does not mention any user confirmation, warning, or clear consent flow before modifying local data. In an agent context, silent persistence of user content can create privacy and integrity risks, especially if users do not realize a casual prompt will update project files automatically.
The eval explicitly instructs the agent to write to persistent storage (book-data.json) without indicating any user notice, confirmation, or consent boundary. In an agent skill, silent persistence can create privacy and integrity risks because user-supplied content may be stored unexpectedly and later reused or exposed in ways the user did not anticipate.
No suspicious patterns detected.
