Best agent skills for LLM evaluation

Braintrust agent skills are readable SKILL.md instruction files that guide coding agents through specific LLM evaluation tasks, helping teams create inspectable datasets, scorers, experiments, analyses, and release policies instead of relying on ad hoc examples or opaque scores. The eight highlighted skills cover planning an evaluation, building labeled datasets, writing and validating scorers, running live experiments, analyzing results, discovering failure modes from traces, and defining CI release gates; each has distinct prerequisites and produces a reviewable artifact. Skills can operate on local files and exports without a Braintrust account, while hosted experiments, traces, datasets, scorers, and validation runs require appropriate project access and API credentials. Users are advised to select only the skill appropriate to their current stage, review its instructions and proposed data access or costs before authorizing execution, and avoid sharing API keys in chat. The skills can be installed individually or as a library through the skills CLI and support compatible coding agents including Claude Code, Cursor, and Codex.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.