Verification Guide

Make the existing harness usable without repeated environment exploration. Apply when onboarding to an unfamiliar verification path or when its setup is repeatedly rediscovered. Update the project's existing guide; a new ticket does not need a new skill or harness.

Discover the Existing Path

Read host workflow/testing docs and inspect actual scripts, fixtures, runner help, and CI. Identify what currently exists, what it covers, and what is still planned. Follow the approved scope and public test interfaces; use accepted journeys and an existing actor's task.

Map the relevant user behavior to the smallest existing representative verification path. Identify prerequisites, setup cost, current owner of shared services, isolated ports/data/accounts/artifacts, and cleanup responsibility. Credentials are references to approved setup, not values to copy into the guide. Missing entrypoints are gaps to report, not commands to invent.

Walk It and Record What Works

Prefer inexpensive discovery and health checks before heavy execution. Run a representative authorized path using actual commands:

• Launch: required setup, existing startup/fixture entrypoint, target build and environment.
• Ready: observable health, identity, data, or geometry predicate that makes the path runnable.
• Drive: public action or test entrypoint and the behavior it verifies, including a relevant failure boundary when needed.
• Evidence: command/exit, source/build identity, actual coverage and result, raw evidence location, and limitations. Preserve a failed attempt as failed.
• Cleanup: owned resources, active consumers/creators, interruption recovery, and reconciliation outcome according to host policy.