The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external collaboration tools, which creates a surface for indirect prompt injection attacks where an attacker could influence the agent's output by sending specific messages in Slack or email.
Ingestion points: The instructions in examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md explicitly direct the agent to read Slack posts, emails, Google Drive documents, and calendar events.
Boundary markers: There are no instructions for the agent to use delimiters (like XML tags) or specific directives to ignore instructions that might be contained within the retrieved documents or messages.
Capability inventory: The skill is primarily focused on text generation and summarization; it does not demonstrate capabilities for executing code or performing sensitive system-level operations.
Sanitization: The skill lacks any defined validation or sanitization logic to filter out potentially malicious content from the ingested data before incorporating it into the final communication draft.

internal-comms