The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes data from external, potentially untrusted sources like Slack, Email, and Google Drive. This creates a surface for indirect prompt injection where malicious instructions embedded in a Slack message or document could influence the agent's behavior.
Ingestion points: examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md instruct the agent to read from Slack, Google Drive, Email, and Calendar.
Boundary markers: No delimiters or specific "ignore embedded instructions" warnings are provided for the data being fetched.
Capability inventory: The agent is expected to use tools for reading communication platforms (Slack, Gmail/Email) and document stores (Google Drive).
Sanitization: No sanitization or validation logic is specified for the content retrieved from these external sources.
[DATA_EXFILTRATION]: The skill encourages the agent to access sensitive information stores, including private emails, Slack communications, and internal documents. Accessing these services is necessary for the skill's functionality but introduces a risk of data exposure if the agent's output is misdirected or if the agent is manipulated via indirect injection.

internal-comms