The Agent Skills Directory

Privilege Escalation (HIGH): The skill exposes administrative functions that can grant 'owner' status (defined as full admin access) without implementing any permission checks. Specifically, the /trust <user> owner and /pairing approve <code> commands in index.ts execute their logic regardless of who is calling the function, enabling unauthorized users to elevate their trust levels.
Indirect Prompt Injection (LOW): The skill is highly vulnerable to instructions embedded in user-provided data.
Ingestion points: The args parameter in the execute function in index.ts accepts raw strings that are parsed into commands.
Boundary markers: Absent. There are no delimiters or developer instructions to ensure the agent ignores embedded commands within the data it processes.
Capability inventory: The skill has the capability to modify the underlying database via svc.setOwner, svc.approveRequest, and svc.removePairedUser (found in index.ts).
Sanitization: Absent. Arguments are processed using basic whitespace splitting (args.trim().split(/\s+/)) with no validation of the identity or authority of the input source.

pairing