Repository navigation
Reports that say what to do. Every ranked function now shows one of four verbs: Done (a test fails when it breaks), Prove (a test runs it; confirm it catches a break), Check (a test file points here; read it first) or Write (no test catches a break yet). The one-page summary leads with what to do for the 20 highest-priority functions. The detailed report is a filterable worklist with a side panel (evidence, why it ranks, two drafted tests, a copyable prompt for your coding agent and the proof command), a map of where tests reach, and run details. Every run also writes .orangepro/tests.csv, one row per drafted test, ready for TestRail, Xray or Zephyr. When nothing changed since the last run, start reuses its reports in about a second (--fresh runs in full). Code that a coverage run executes now ranks as detectable as code a test calls (ranking v3). Fixes rankings that could vary with machine load on large repositories. New README.