The Agent Skills Directory

[SAFE]: The skill is designed for the specific purpose of analyzing experiment data from Datadog LLMObs via official tools. All logic and tool usage identified align with this legitimate analytical function.
[DATA_EXFILTRATION]: The skill generates deep-links to the official Datadog application domain ('app.datadoghq.com'). These links are used to direct users to comparison views in the UI, which is a standard feature of the vendor's platform and involves a well-known service.
[EXTERNAL_DOWNLOADS]: Includes documentation containing an MCP server URL ('mcp.datadoghq.com'). This is provided as an instruction for users to manually configure their environment with official vendor tooling and does not constitute a silent or malicious download.
[PROMPT_INJECTION]: The skill processes experiment data which may include untrusted model inputs and outputs. While this represents a potential surface for indirect prompt injection, the skill is limited to report generation and explicitly requires PII redaction in its output instructions, maintaining a safe posture for processing external content.

experiment-analyzer