The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill is designed to ingest and process data from various untrusted internal sources, which is a significant vulnerability surface for indirect prompt injection.
Ingestion points: examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md specify reading from Slack, Google Drive, Email, and Calendar.
Boundary markers: Absent. The instructions do not define delimiters or provide warnings to ignore commands within processed content.
Capability inventory: While no scripts are provided, the skill intended for the agent involves reading sensitive enterprise data and drafting/sending communications, establishing a pathway for potential manipulation via malicious data.
Sanitization: Absent. There is no instruction to escape, validate, or filter the content retrieved from these external tools.

internal-comms