The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because several of its core 'loops' are designed to ingest untrusted data from external sources.
Ingestion points: As detailed in references/loop-catalog.md, the agent is instructed to scan external platforms including news sites, Reddit, X (Twitter), and community forums (Slack, Discord) for marketing opportunities and social listening.
Boundary markers: The skill does not define specific text delimiters for external data but addresses this risk by mandating a human-in-the-loop requirement for all Tier-2 (gated) actions, such as sending emails or posting to social media.
Capability inventory: The agent can draft social media replies, prepare PR pitches, and generate email content based on external inputs, which creates a path for indirect influence on agent output.
Sanitization: The references/loop-guardrails.md file provides comprehensive guidelines for PII handling and compliance (GDPR/CAN-SPAM) but does not include technical sanitization methods to filter potential instructions embedded in the ingested data.

marketing-loops