Check the report behind PASS
Compare the command result with the actual report
A wrapper can finish successfully and still leave you without a usable report. This check compares the actual command result, the recorded status and the report’s fields.
The corrected Frequency wrapper now returns failed command exits. It still prints PASS when an exit-zero verifier produces no report or a bad one. The reader catches those cases and keeps each raw result.
Run the example, compare another wrapper, or keep the check in your CI. Bring your output to the shared case thread.
The retained run covers 14 cases and 63 tests. Failed commands, absent and malformed reports, wrong inputs, stale times and a linked report stay separate. These are injected commands and example reports; they do not run the producer’s verifier.
Credit stays with imokokok for the original finding, Pico/Håkon Åmdal for the outside E030 run and kept check, and the native producers for their contracts. The original vector and outside result remain unchanged.
Source, run contract and retained results
The wrapper is pinned at cf7389097fe3a404b3557da2e72fdd8cbe962b81. The source map names every input and owner. The contract separates an observed command result from a producer-declared report time.
Qualification, test results, case results and integration retain the exact preparation source and checks.
Operator identity is caller-declared. The native report has no invocation token. A copied report with a forged current time cannot establish fresh execution.