The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill involves fetching and processing content from external, untrusted sources such as competitor websites, social media threads, and customer feedback.
Ingestion points: Data enters the agent's context through various monitoring loops described in references/loop-catalog.md, references/outbound-operator.md, and references/seo-operator.md.
Boundary markers: The instructions provide clear warnings that fetched content is untrusted and explicitly forbid the agent from following instructions embedded within that data.
Capability inventory: The skill manages workflows that primarily focus on analysis and drafting; all sensitive actions (spending, sending, publishing) are classified as Tier-2 and require a human checkpoint.
Sanitization: The skill mandates summarizing or diffing content rather than direct interpolation and includes veto lists for filtering sensitive or unsafe topics.
[SAFE]: The skill includes comprehensive security best practices, including PII protection (recommending hashing/IDs instead of raw data), a two-tier action model for risk management, and a requirement for a documented manual kill switch for all scheduled loops.
[EXTERNAL_DOWNLOADS]: The skill references an external documentation file hosted on GitHub (github.com/coreyhaines31/marketingskills). This is a reference to a markdown file for instructional purposes and does not involve automated script execution or package installation.

marketing-loops