The Agent Skills Directory

[SAFE]: The skill implements a comprehensive safety framework for recurring tasks (loops). It explicitly defines 'Gated' actions (Tier 2) such as spending money, shifting budgets, and publishing content, which require mandatory human checkpoints unless otherwise authorized with strict caps.
[SAFE]: Data privacy is addressed in the references/loop-state.md file, which instructs the agent to avoid logging raw PII in state or run logs, instead recommending the use of internal IDs or hashes.
[SAFE]: The skill incorporates robust idempotency patterns (cursors, dedupe sets, and cooldowns) to prevent the agent from repeating actions, double-contacting users, or performing redundant tasks.
[SAFE]: For activities involving external data (e.g., social listening or newsjacking), the skill includes 'Veto lists' and 'Always-escalate' lists to prevent the agent from acting on sensitive, tragic, or crisis-related topics without human oversight.
[SAFE]: Scheduling is handled through native platform tools (CronCreate, ScheduleWakeup) or standard cron jobs, with no evidence of unauthorized persistence or backdoors.

marketing-loops