← ClaudeAtlas

ck-verify-createlisted

Establish a repeatable project verification method from real user paths, reusing existing checks and creating a skill only for a useful gap. Use when verification knowledge is missing or hard to reuse for an app, CLI, library, service, or analysis workflow.
Wang-Cankun/cankun-skills · ★ 2 · AI & Automation · score 78
Install: claude install-skill Wang-Cankun/cankun-skills
# Create a verification skill Give the project a repeatable way to exercise its real interface and prove behavior. First determine what is missing from the existing verification method. When a skill is useful, write it for an agent arriving cold, mid-task. `ck-verify-<project>` is a naming pattern: replace `<project>` with a concrete project slug, such as `ck-verify-count-summary`. Keep the generator reusable. Project commands, environment paths, reference versions, fixtures, and tolerances belong in its generated output, discovered from that project's evidence. ## 1. Identify the useful gap Inspect the codebase and ask only for facts or judgments that cannot be found there: - **Surface:** what does a user actually touch? Web or desktop UI, CLI/TUI, API, library calls, notebook, or workflow command? Identify the primary path and other supported entry points. - **Run:** which project commands prepare and run it? Prefer documented scripts and declared environments. Identify readiness, required inputs, services, authentication, reference data, and expected resource cost. - **Drive:** which existing harness can exercise that path? Reuse browser automation, CLI fixtures, PTY helpers, HTTP clients, integration tests, workflow test profiles, or public library examples. Choose available tools; neither an agent platform nor a particular connector is required. - **Observe:** which meaningful results establish behavior? Consider screenshots, response values, terminal transcripts, s