The Agent Skills Directory

[SAFE]: The skill establishes a structured environment for AI agents, mandating risk assessment via specific thinking loops and deterministic execution protocols.
[SAFE]: Security hygiene is explicitly addressed, with instructions to treat external web content as hostile and to redact secrets and PII before sharing context.
[SAFE]: The skill enforces least-privilege access by restricting agents to defined tool allowlists and mediating terminal executions through a proxy interface.
[SAFE]: Process integrity is maintained through mandatory context verification and the use of persistent audit logs for task completion.

experiment