The Agent Skills Directory

[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by design. It instructs the agent to read and summarize data from various corporate communication channels which may contain adversarial instructions.
Ingestion points: The agent is directed to gather information from Slack messages, Google Drive documents, emails, calendar events, and external press releases (examples/3p-updates.md, examples/company-newsletter.md, examples/faq-answers.md).
Boundary markers: The instructions do not specify the use of delimiters or provide "ignore embedded instructions" warnings for the external data it processes.
Capability inventory: The skill's primary capabilities involve reading diverse corporate data sources and writing formatted text updates based on that content.
Sanitization: The skill lacks instructions for sanitizing, escaping, or validating the content retrieved from external sources before interpolation into the agent's output.
[DATA_EXFILTRATION]: The skill requires broad access to sensitive corporate information including emails, chat logs, and internal documents to function. While this is consistent with the skill's stated purpose, it represents a significant data exposure surface. However, the analysis found no evidence of exfiltration attempts to external domains or untrusted third parties.

internal-comms