The Agent Skills Directory

[SAFE]: The skill implements standard analytical logic for meeting notes processing. No suspicious command execution, network activity, or credential theft patterns were found.
[DATA_EXFILTRATION]: The skill references a local file at 'content/org/colleagues.json' for the purpose of verifying participant names. This operation is performed locally within the file system and does not involve any external network calls or data transmission.
[PROMPT_INJECTION]: The skill processes untrusted external data (meeting transcripts). While this creates a surface for indirect prompt injection, the skill's instructions are confined to classification and attribution heuristics, and it lacks dangerous capabilities (such as code execution or system modifications) that could be leveraged by an injection attack.

analyzing-meetings