The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to ingest data from untrusted external sources, including web searches, community forums, and user discussions (SKILL.md). This creates a surface for indirect prompt injection, where malicious content embedded in these sources could influence the agent's output or logic. There are no explicit boundary markers or sanitization steps defined to isolate these inputs.- [DATA_EXFILTRATION]: The framework directs the agent to aggregate sensitive organizational information, such as CRM records, email threads, and internal wikis (SKILL.md). While the skill includes guidelines for confidence and escalation, the process of synthesizing this data for potential delivery to customers involves a risk of exposing internal-only context to external parties.

customer-intel