docs: prefer rubrics over llm-grader with inline prompt#1097
Merged
Conversation
Replace all instances of the llm-grader-with-inline-prompt antipattern in examples, skill docs, and web docs with rubrics assertions. Rubrics give per-criterion score breakdowns and make failure debugging easier; inline llm-grader prompts are an opaque free-form alternative. Also update agent-eval-layers.mdx tables to reference `rubrics` instead of `llm-grader` for LLM-based quality checks, and delete two stale baseline files (need regeneration after grader type change). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Deploying agentv with
|
| Latest commit: |
2f63c97
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://42ae0559.agentv.pages.dev |
| Branch Preview URL: | https://docs-prefer-rubrics-over-inl.agentv.pages.dev |
Replace full `type: rubrics` + `criteria:` blocks with inline string shorthand where there are no weights, required gates, or composite naming concerns. Composite children in threshold-evaluator keep named rubrics graders to preserve threshold aggregation semantics. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Replaces all instances of the
llm-grader-with-inline-prompt:antipattern across examples, skill docs, and web docs withrubricsassertions.Why:
rubricsgives structured, per-criterion scores that are easier to debug when a test fails. An inlineprompt:on anllm-graderis an opaque free-form blob that produces the same result but with no criterion breakdown.Preferred hierarchy (now reflected in examples):
contains,regex,equals,is-json) — no LLM cost, fully reproduciblerubrics/ string shorthand — structured LLM judgment with per-criterion scoresllm-grader(no prompt) — implicit eval againstcriteria, fine for simple casesllm-graderwithprompt: ./file.md— custom prompts; the legitimate use case for a file-backed graderFiles changed:
examples/features/preprocessors/evals/dataset.eval.yamlexamples/features/threshold-evaluator/evals/dataset.eval.yamlexamples/features/experiments/evals/coding-ability.eval.yamlexamples/features/default-evaluators/evals/dataset.eval.yamlplugins/agentv-dev/skills/agentv-bench/SKILL.mdapps/web/src/content/docs/docs/evaluation/examples.mdxapps/web/src/content/docs/docs/guides/agent-eval-layers.mdxDeleted baselines (need regeneration with real LLM calls):
examples/features/threshold-evaluator/evals/dataset.eval.baseline.jsonlexamples/features/default-evaluators/evals/dataset.eval.baseline.jsonlNot changed:
apps/cli/test/commands/eval/pipeline/fixtures/input-test.eval.yaml— this test fixture explicitly testsllm-graderpipeline behavior (prompt content writing,llm_grader_results/directory) and is not a teaching example.🤖 Generated with Claude Code