Security audit
Security checks for vulnerabilities and agentic risk
This skill only restructures meeting notes the user provides, with privacy risks that are expected for that purpose and should be reviewed before sharing outputs.
Install only if you are comfortable pasting meeting content into the agent. Redact secrets, credentials, regulated data, and unnecessary personal details, and review generated emails or summaries before sharing them with clients, coworkers, or wider groups.
The skill explicitly instructs the model to list every attendee by name and include role details where possible, which can reproduce personally identifiable or sensitive organisational information from pasted notes. In contexts involving confidential meetings, this increases the chance of unnecessary data exposure in generated summaries or downstream sharing.
Generating one self-contained follow-up email per attendee encourages the model to restate meeting context, decisions, and actions multiple times, potentially spreading sensitive information beyond what each recipient needs to know. Because the emails are designed to be copy-paste ready, the skill increases the risk of over-disclosure if users send them without carefully reviewing audience scope.
The private 1:1 notes mode asks the model to extract feedback, development goals, and mood/energy signals, which can surface sensitive employee or HR-related information from raw notes. Even though the section is labeled private, the skill still structures and amplifies sensitive personal data, making accidental retention, sharing, or misuse more likely.
The skill explicitly encourages users to paste raw meeting notes, transcripts, Slack threads, and handwritten-note descriptions, which commonly contain confidential business information, personal data, or commercially sensitive decisions. Without any warning, minimization guidance, or handling expectations, users may disclose sensitive content to the system unintentionally, making this a real privacy and data-exposure risk.
Line L417 instructs the skill to use 'UK English by default,' which imposes a locale preference in natural-language behavior. The file does not offer an opt-in or user choice for language/locale, so this is a policy concern under the language/locale rule.
This is a natural-language locale policy issue because the skill description specifies a particular regional language variant as the default behavior with no opt-in or alternative. Users outside that locale may receive output in an unwanted format or tone.
No suspicious patterns detected.
