The Agent Skills Directory

[SAFE]: The skill establishes a disciplined operational model for AI agents, prioritizing human oversight and idempotent execution to ensure stability and safety across automated marketing workflows.\n- [PROMPT_INJECTION]: The skill processes untrusted data from external sources (news feeds, social media, search reports), presenting a potential surface for indirect prompt injection. This is addressed through a mandatory two-tier action model where all high-risk actions (spending, publishing, sending) require explicit human approval.\n
Ingestion points: references/loop-catalog.md (newsjacking, social-listening, and voice-of-customer loops).\n
Boundary markers: references/loop-guardrails.md defines a strict boundary between autonomous-safe (Tier 1) and gated (Tier 2) actions.\n
Capability inventory: Orchestrates various marketing tools via separate skills for data retrieval and content drafting.\n
Sanitization: Includes 'veto lists' for filtering sensitive or inappropriate external content in the newsjacking loop and 'self-checks' to validate signals before acting.

marketing-loops