The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow where the agent reads and summarizes content from untrusted external sources, creating a significant attack surface.
Ingestion points: Files examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md instruct the agent to retrieve information from Slack messages, shared Google Drive documents, and company emails.
Boundary markers: The instructions do not define any delimiters or provide guidance to the agent to distinguish between the skill's instructions and potentially malicious instructions embedded within the ingested messages or documents.
Capability inventory: The skill is intended to operate with read access to Slack, Google Drive, Email, and Calendar tools.
Sanitization: There are no mechanisms described to sanitize, validate, or filter content retrieved from corporate communication tools before it is processed and presented to the user.

internal-comms