The Agent Skills Directory

[DATA_EXPOSURE]: The skill's workflow involves accessing and interpreting data from internal organizational sources such as production databases, Google Analytics, Datadog, and Zendesk. These operations are intended to provide evidence for root cause analysis and are performed through standard agent-tool interactions based on user-provided data or tool access.
[INDIRECT_PROMPT_INJECTION]: The external-check-agent and verdict-agent ingest data from external web searches and user-supplied reports. This creates an inherent attack surface where malicious instructions embedded in web content or processed files could attempt to influence the agent's logic. However, the skill provides structured templates and quality gates (via the critic-agent) to mitigate such risks.
[COMMAND_EXECUTION]: The orchestration logic defines a sequence of sub-agent dispatches. While the skill suggests database queries (e.g., SQL SELECT statements), these are presented as examples for data gathering rather than being executed directly as arbitrary system commands by the skill itself.

problem-analysis