The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and displays content from emails, creating a potential surface for indirect prompt injection if an email contains instructions meant to override agent behavior.
Ingestion points: The skill ingests data from external emails through the list_emails and search_emails tools as described in the invocation steps in SKILL.md.
Boundary markers: There are no explicit boundary markers, delimiters, or instructions provided to the agent to treat retrieved email content as data only and disregard potential commands within it.
Capability inventory: This skill's logic is restricted to data aggregation and display; no high-risk capabilities such as file system writes, external network requests, or shell command execution are present in the skill body.
Sanitization: The skill does not describe any specific sanitization, filtering, or validation processes for the email metadata before it is interpolated into the agent's context.

stats