agentsclimarketplace

Evaluate with braintrust

Skill hiteshbandhu/skills-i-use/skills/ai-engineer-talks/evaluate-with-braintrust

Evaluates LLM and agent products with Braintrust—offline/online evals, SDK/CI workflows, eval platform design, Zapier/Notion product ops, and Loop-style optimization. Use when the user mentions Braintrust, eval playground, autoevals, production log evals, or agent quality platforms.From its SKILL.md

Install
npx -y skills add hiteshbandhu/skills-i-use --skill evaluate-with-braintrust

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 2 commands, including `cp -r skills/evaluate-with-braintrust ~/.cursor/skills/` and 1 more.

SKILL.md

1.9 KB, 428 tokens by cl100k_base, as published. Nobody here has run it

Evaluate with Braintrust

Action playbook from seven Braintrust @ AI Engineer talks. Do not summarize — pick a workflow.

Supporting files: workflows.md · source-index.md

Related skill: run-llm-evals (cross-vendor eval theory).

Optional: ./skill-outputs/evaluate-with-braintrust/


Step 0 — Pick workflow

What is the user trying to do?
├─ Eval fundamentals (datasets, scorers, modes)        → evals-101
├─ Build/buy eval platform architecture                → platform-design
├─ Five lessons / future of evals (velocity, Loop)     → eval-ops
├─ Zapier-style product + eval integration             → product-ops
├─ Notion-class world-class AI quality bar             → world-class-products
└─ Complex app workshop (Trainline patterns)           → complex-apps

Install

cp -r skills/evaluate-with-braintrust ~/.cursor/skills/
cp -r skills/evaluate-with-braintrust ~/.codex/skills/

Source: playlists/braintrust-ai-engineer/.


Cross-cutting rules

RuleSource
Eval = data + task + scorers; offline and online[src-001 @ 0:07:12]
Prod logs must feed datasets and online evals[src-001 @ 0:13:13]
Unify offline experiments with trace replay[src-002 @ 0:14:15]
24h model swaps when evals healthy (Notion bar)[src-003 @ 0:40]
Human-in-loop on Loop optimizations[src-004 @ 0:03:33]

Output

Name workflow; artifacts under ./skill-outputs/evaluate-with-braintrust/ when requested.

What ships with it: 3 files

4.5 KB alongside SKILL.md

Gives 0 of the 12 instructions most evals benchmarks skills give in 428 tokens

Counted across 499 of the 513 authors here whose files we hold, read 2026-09-06

  • Spawn with-skill and baseline runs in the same turnin 31 of 499, across 24 files
  • Keep SKILL.md under 500 linesin 31 of 499, across 24 files
  • Draft assertions while test runs are in progressin 31 of 499, across 24 files
  • Compare against the baseline after changesin 31 of 499, across 13 files
  • Define evals before codingin 26 of 499, across 17 files
  • Run evals frequently during developmentin 25 of 499, across 16 files
  • Keep evals fastin 24 of 499, across 15 files
  • Version evals with codein 24 of 499, across 15 files
  • Generate the eval viewer before evaluating outputs yourselfin 24 of 499, across 17 files
  • Generate an eval report after runsin 24 of 499, across 15 files
  • Track pass@k metrics over timein 22 of 499, across 14 files
  • Save a baseline before making changesin 21 of 499, across 9 files

Said here and by no other author read

  • Name the chosen workflow and what you are producing
  • pick one workflow instead of summarizing
  • match the user's goal to a workflow
  • structure evals as data, task, and scorers
  • feed production logs into datasets and online evals
  • unify offline experiments using trace replay

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.