Andrew Slutsky

release: 3.11.8 — version rev at DRU's word; no functional changes from 3.11.7

fix: a present-but-empty interpreter is refused, not defaulted -- the or-chain over job['interpreter'] silently substituted the platform default for [], defeating the non-empty check beneath it (the job ran; a shell exiting 0 recorded success). Only an absent interpreter falls back now. Sibling test pins [] alongside the string case. Also: the intake fuzz tests write generated text with an explicit utf-8/surrogateescape encoding (the platform default on Windows, cp1252, cannot encode what Hyp...

tests: the seeded-workflow tests assert the workflow as text -- CI carries no YAML library (the pinned requirements-ci.txt has none), so the module's yaml import broke the tag run's test job while the local venv masked it

release: 3.11.7 — strictness batch, claims-audit doc corrections, seeded mutation pipeline

ci: seeded mutation pipeline — a scheduled, non-gating twin of the release mutation job (seed = run number or a dispatched value, printed in the log and summary; PYTHONHASHSEED + HP_MUTATION_SEED carry it; conftest wires the seed into Hypothesis only when set)

docs: FUTURE — in-memory fuzzing and the LLM's place in a fuzz loop (DRU's entry, with his adoption precondition)

docs: park corrections — the av detections sense and the Windows WFP cage are not built

docs: BC-1 — CONTRIBUTING chat privacy rests on redact-before-write, not a no-transcript guarantee