The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill creates a significant attack surface for indirect prompt injection by directing the agent to process content from various untrusted organizational data sources.
Ingestion points: The files examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md instruct the agent to gather and summarize information from Slack threads, company emails, Google Drive documents, and Calendar events.
Boundary markers: The skill provides no delimiters or explicit instructions to help the agent distinguish its operational guidelines from the potentially adversarial content found in the ingested communication data.
Capability inventory: The agent possesses the capability to read sensitive organizational information and generate reports or announcements that are distributed widely across the company.
Sanitization: There are no mentioned requirements or mechanisms for sanitizing, validating, or filtering external content before it is processed by the model, allowing for potential manipulation of agent behavior through hidden data patterns.

internal-comms