The Agent Skills Directory

[SAFE]: The skill incorporates a robust safety framework defined in 'references/loop-guardrails.md', which establishes a clear distinction between autonomous-safe tasks and gated actions. Any operation involving financial expenditure, content publishing, or external communication is classified as 'Tier 2' and requires a human checkpoint by default.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from external, untrusted sources such as competitor websites, social media feeds, and customer reviews. Ingestion points: Data is gathered from various external triggers including search engine results, social mentions (HN, Reddit, X), and feedback logs. Boundary markers: The skill enforces strict 'self-check' and 'human checkpoint' protocols to verify the validity of external signals before any action is taken. Capability inventory: The system can orchestrate tools for email delivery, social media management, and ad campaign modification. Sanitization: All newsjacking activities are subject to a specific 'veto list' to filter out inappropriate or sensitive topics, and all public-facing output requires manual verification.
[COMMAND_EXECUTION]: As described in 'SKILL.md', the skill utilizes platform-specific automation primitives like 'CronCreate' and 'ScheduleWakeup' to manage the execution cadence of the marketing loops. These tools are used within their intended functional scope to facilitate recurring analysis tasks.

marketing-loops