Generate eval.md evaluation criteria with test cases and run instructions.
npx skills add https://github.com/tya5/reyn --skill eval-builder
Please help me install this Agent Skill. Skill: eval_builder Source: https://github.com/tya5/reyn/tree/main/src/reyn/stdlib/skills/eval_builder Command: npx skills add https://github.com/tya5/reyn --skill eval-builder
It automates the creation of detailed evaluation specifications (eval.md) that assess a skill’s correctness and quality.
Invoke the eval_builder skill to generate an eval specification for a given skill name.
High-intent search queries and answers about installing and using this skill.
To generate evaluation criteria for testing skill performance, you can use an automated eval builder that creates a detailed eval.md document containing specific test cases, quality standards, and run instructions for objective assessment.
An eval.md specification is a document containing evidence-bound quality criteria and test cases for assessing skill correctness. You need one when you want to objectively measure and verify skill behavior across typical and edge scenarios.
You can create test cases for edge scenarios by designing an evaluation specification that probes typical, edge, and failure conditions, producing observable, criterion-based verification of skill behavior and referencing existing artifacts when available.
Yes, you can evaluate skill outputs without manual standards by automating criteria generation. The tool designs specific, evidence-bound quality criteria for each workflow phase, enabling consistent performance measurement without manual specification writing.
The best way to structure run instructions for skill assessment is to align them with workflow phases and quality criteria within a comprehensive eval.md document, ensuring consistent performance measurement and observable verification of skill behavior.
understudylabsGenerate harness, metric, splits, and baseline artifacts from workloads.
NicolasPadovamDevelop, evaluate, and optimize custom agent skills within the Opensquad framework.
imbue-aiGenerates and runs frontend integration tests for Sculptor using FakeClaude.
DTMC-marketplaceAutomate AI system performance evaluation with Deepeval and generate markdown reports.
