The Agent Skills Directory

[SAFE]: The skill uses established internal tools (search, employee_search, meeting_lookup, and chat) to collect information relevant to its stated purpose of onboarding. These tools are associated with the vendor's environment and are used for standard information retrieval tasks.
[PROMPT_INJECTION]: The skill processes untrusted data from internal document searches and meeting summaries, which represents a potential surface for indirect prompt injection.
Ingestion points: Data enters the context via search and meeting_lookup results in Phase 3.
Boundary markers: The instructions do not specify explicit delimiters or "ignore" instructions for the data retrieved from search results.
Capability inventory: The skill's capabilities are limited to information retrieval and synthesis; it lacks dangerous actions such as writing to the file system, making external network requests, or executing shell commands.
Sanitization: The skill includes a vetting phase (Phase 4) where the agent is instructed to evaluate the accuracy, freshness, and essentiality of the content, which serves as a logical filter for the retrieved data.

onboarding