Security audit
Security checks for vulnerabilities and agentic risk
The skill is a coherent rental advisor, but it asks the agent to run broad live searches and even use shell-based curl scraping without clear limits.
Review this skill before installing if your agent has shell, web, profile-memory, or alerting tools enabled. It appears aimed at legitimate rental help, but you should restrict it to user-requested searches, prefer bounded web/search tools over exec + curl, and avoid storing rent/profile data or creating alerts unless the user explicitly agrees.
The top-level description is broad enough to activate on many ordinary housing conversations, causing the skill to engage outside a narrowly intended scope. Overbroad triggering increases the chance of unintended tool use, unnecessary data collection, or the model following this skill when a simpler non-tool response would suffice.
The Area Recommendation trigger criteria are scenario-based and open-ended, so the skill may activate for vague location discussions that do not require this workflow. In context, this matters because the skill then instructs requirement collection and potential follow-on searches, expanding data handling and action-taking from a loosely matched prompt.
The Property Search triggers are especially broad for a function that can perform external searches and collect listing/contact details. Because this function can reach external sources and handle current market data, ambiguous activation creates a higher risk of unnecessary scraping, privacy issues, or unsafe tool use compared with purely advisory behavior.
Phrases such as 'Asks what to watch out for when signing' and 'Worried about contract traps' are broad, everyday expressions that could match many generic legal or consumer questions. The section does not provide enough specificity or exclusions to prevent accidental activation.
The skill explicitly authorizes exec + curl for web scraping, which expands the capability from bounded information retrieval to general shell execution. In an agent setting, this can enable command execution, unsafe network access, or abuse of user-influenced inputs if later combined with dynamic parameters, making the skill materially riskier than a normal rental-advice workflow requires.
Statements like 'Whether rent in a certain area/community is reasonable' and 'Reasonable rent range for certain conditions' are intent descriptions, not precise triggers. Without examples of non-triggering cases or stricter scope, the activation boundary remains ambiguous.
The document presents dispute-resolution channels and legal resources entirely in one locale/jurisdictional framing, including a specific government hotline, without indicating that these resources may vary by user region. This can amount to a language/locale policy issue because it implicitly assumes a specific locale rather than offering a choice or clarifying the intended jurisdiction.
