Security audit
Security checks for vulnerabilities and agentic risk
This is a coding workflow guide that only stores explicitly requested local preferences and contains no executable code or network behavior.
Before installing, be aware that the skill can remember coding preferences in ~/code/memory.md when you explicitly ask it to, and those notes remain on local disk until you edit or delete them. Avoid asking it to save sensitive personal, credential, or project-secret information.
The guidance says to plan when the user says "build", "create", or "implement", which are very common verbs used in many ordinary requests. Without additional constraints or negative examples, this trigger description is broad enough that the skill could activate in situations where planning is unnecessary or unintended.
The skill instructs the agent to persist user preferences to a specific local file path (~/code/memory.md), but it does not clearly warn the user at the point of use that information will be written to local storage. Even though the criteria require explicit user request and a confirmation question, the persistence behavior could still surprise users and lead to unintended retention of potentially sensitive preference data.
No suspicious patterns detected.
