# Selected validation recipe, job986 Frozen finite claim:16 new oldT97 first66-slot prefixes,31 fixed within-window placements each,7 counting-silent windows,0 strict ordinal-template certificates,0 strict controls, H0. All capacities/phase maximizers and prefix/seed coverage are complete for this registered domain. No population/randomization inference or source optimization claim. Reconstruct the flattened manifest into a clean directory and verify every listed SHA. CPython3.12.13/POSIX resource module was observed; no third-party Python dependencies. Run: python3 transfer986-checker.py transfer986-input.json transfer986-result.json Expected exact stdout: {"H":0,"capacity_vectors_checked":528,"complete":true,"controls_checked":496,"eligible":7,"passed":true,"template_strict":0,"windows":16} plus one final newline, exit0. The checker consumes the published producer target and frozen input, reconstructs old admissibility with gcd, enumerates every phase/slot capacity, independently regenerates the seeded31 distinct placements perwindow, and checks every value/maximizer/H. It does not validate370's source optimizer or historical capacity table. Frozen input defines the exact vector under test; source text/hash is separately supplied for attribution inspection. Observed3.102CPU sec for the full checker, peakRSS below24MB on the stated host. Prospective validation cap20CPU seconds incl four controls,256MBRAM/100MBdisk; separate judgment15minutes. Scripts require POSIX resource/setrlimit. A platform lacking it should report capability inability, preserving the original bytes, no shim-pass claim. In separate copies of the target, remove one window; add1 to the first control's first weight; add1 to the first templateC; set the first control's first101-prime maximizer to101. The unchanged checker must reject each with missing window, weight multiset, template capacities or phase domain respectively. These4 controls were actually executed and rejected, with stderr/exit observations in transfer986-controls.out. Keep the original producer result unchanged. To replay discovery separately, run the supplied producer with transfer986-input.json and compare its stdout bytewise to transfer986-result.json. This is optional reconstruction, not needed to check the target. The published local-run helper uses outputs/ and work/statistic-986 relative paths; create those directories and place files underoutputs to replay its actual measurement/control workflow. Timings and resource observations vary and are not deterministic reproduction targets. No source-support capacity/70-prefix or14N52-branch rerun is part of this check. All16 disjoint10000 scans at a9409+20000j,j1..16 and all66 slots/19q-domain phases in528 vectors are decisive coverage for the finiteH statement. No claim outside the frozen sample. Independent algorithm is the author checker, not an independent contributor review; worker methodrerun should state shared code/input/spec and visible expected answer. No observed check_receipt is invented here.