Measure whether a server's live tests actually RUN — a cron-triggered workflow whose pytest call selects the `live` marker and whose failure reaches somebody — instead of only being marked and excluded from CI with `-m "not live"`. LIVE_UNSCHEDULED when…
Compare the field names a server's code reads out of a response against the names the source delivers live — FIELD_CASE_DRIFT when the column exists under a different spelling, FIELD_MISSING when it is gone, UNVERIFIED when the source, the site or the mapping…
Measure whether a column the code sends through int()/float() actually holds numbers in the live source, and report the measured share that does not — VALUE_DOMAIN_DRIFT with the proportion in the finding, UNVERIFIED when nothing could be concluded. Catches…
Compare the code a skill repository ships under reference/ against the servers that copied it, in BOTH directions — REFERENCE_STALE when the template is behind the servers, REFERENCE_UNADOPTED when a server is behind the template, UNVERIFIED when the mapping…
Check whether the dependency bound declared in pyproject.toml is also in force in the lockfile the deployment installs from. Reports LOCK_DRIFT with both diverging specifiers, LOCK_UNSATISFIED when the pinned version violates the declaration, LOCK_STALE from…
Verify that every identifier the documentation cites — rubric codes, check IDs, constants, file paths, collection membership — actually resolves in the code. Catches a README or SECURITY.md that describes checks and constants which do not exist.…
Compare a bilingual documentation pair (README.md/README.de.md, SECURITY.md/SECURITY.de.md) for structural parity — heading skeleton, per-section bullet counts, tagged code blocks, link targets, and how far the translation lags in git. Never compares prose,…
Measure what the package on PyPI actually does — install it into a throwaway venv, read the User-Agent it puts on the wire, check that every module imports, start its console script, and check its declared dependency ranges for a missing upper bound. Use when…