The Agent Skills Directory

[SAFE]: The skill provides legitimate instructions and code examples for the Arize Phoenix framework, matching its stated purpose of building evaluators for AI applications.- [EXTERNAL_DOWNLOADS]: Installation instructions target official Arize Phoenix packages and reputable AI provider SDKs. These are standard and well-known industry dependencies.- [CREDENTIALS_UNSAFE]: The skill follows safe practices by using placeholders for API keys and credentials, avoiding hardcoded secrets.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted LLM outputs as part of its evaluation workflow. It utilizes variable interpolation within prompt templates, which is a standard pattern for judge models. The documentation promotes the use of XML tags for structural data separation.

phoenix-evals