Analyze a target SKILL.md and generate an initial evals.json with test scenarios.
npx skills add https://github.com/matantsach/snapeval --skill create-evals
Please help me install this Agent Skill. Skill: create-evals Source: https://github.com/matantsach/snapeval/tree/main/skills/create-evals Command: npx skills add https://github.com/matantsach/snapeval --skill create-evals
Onboarding new skills into an eval workflow can be time-consuming. This skill automates the initiation by analyzing the target SKILL.md, outlining required evals.json structure, and guiding through a phased setup.
Run this guided flow to generate an initial evals.json and a benchmark plan for a new skill.
High-intent search queries and answers about installing and using this skill.
To generate an evals.json file, the skill analyzes the target SKILL.md to map core behaviors and input dimensions, then creates 5-8 test scenarios covering happy paths, edge cases, and negative tests before writing the configuration.
The best way to structure skill evaluation scenarios is to map core behaviors from SKILL.md and design 5-8 targeted tests covering happy paths, edge cases, and negative tests to validate the skill comprehensively.
You do not need an existing evals.json to start benchmarking; if the file does not exist, the system generates it under evals/evals.json and prepares a safe execution plan for the initial benchmark run.
You can automate skill evaluation onboarding by reading the target SKILL.md, asking 1-3 targeted clarifying questions to identify gaps, and automatically producing the initial evals.json configuration alongside a benchmark execution plan.
Before generating an evals.json configuration, 1-3 targeted clarifying questions are asked to resolve gaps identified during the SKILL.md analysis, ensuring the test scenarios accurately validate the intended core behaviors.
hulusi-tuncMap core tasks across ability spectrums into a task × ability matrix with prioritized gaps.
heyitsnoahCreate, refine, and evaluate Claude skills with SKILL.md frontmatter.
quanhavnPlan, track, and execute software work using project context and tracks.
felipesantiago-coderCreate AI skills with structured evaluation and benchmarking workflows for GLM Code.
