Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local CRM assistant that stores sales pipeline data in plaintext JSON, with no evidence of remote execution, exfiltration, or hidden behavior.
Install only if you are comfortable keeping CRM records in local plaintext JSON files. Use clear CRM-oriented commands, review confirmations carefully, and keep the data directory out of public repositories or shared backups if it contains prospect, contact, or revenue information.
The README uses broad natural-language examples such as adding leads, logging proposals, and setting reminders without defining clear trigger boundaries or exclusions. In an agentic environment, overly general conversational triggers can cause unintended activation from ordinary discussion or pasted text, leading to accidental creation, modification, or closure of CRM records.
The skill instructs the agent to parse ordinary natural-language mentions of leads, deals, contacts, and sales activities into structured CRM actions, which creates an overly broad activation surface. In a chat environment, routine discussion about sales could be misinterpreted as a command, causing unintended reads or writes to the local CRM database and accidental persistence of sensitive business information.
The skill encourages users to enter names, companies, deal values, follow-up details, and pipeline status, but the description does not clearly warn that this constitutes storage of potentially sensitive business and personal data. Even with a 'local-first' claim, users may unknowingly expose prospect information, revenue figures, or confidential sales activity to local logs, backups, or shared devices.
No suspicious patterns detected.
