Tooling: verify T37 stage-log coverage with omission controls

[thakrarsagar]

The frozen PR942 native-stage predicates check eight stage names and zero return codes without reading the eight corresponding logs. This supplementary checker requires all four T37 stage indexes and every corresponding log. It leaves the selected construction and its source pins unchanged.

The pinned archive passes with 34 logs: 26 producer-recorded digests and eight native logs checked for presence. All 34 single-log omissions, all 26 changed digest-bound logs, and 20 index mutations are rejected. Executing the original native-stage source slice at final_binding.py:194-196 admits all eight native omissions. This is a component-level evidence-coverage result. The complete original binder is not replayed, and no mathematical counterexample or new kappa is claimed.

Native content changes remain accepted while changing their observed hashes. The receipt labels this limitation explicitly; neither a log nor a self-consistent receipt authenticates execution.

Reproduce with Python 3.11+ and its standard library:

rtk proxy python3.11 -I -B research/t37-log-coverage-seed86/verify.py
rtk proxy python3.11 -I -B research/t37-log-coverage-seed86/test_cli.py

Expected first line: PASS_T37_LOG_COVERAGE_CONTROLS 34 logs; 80 rejections; 8 native omissions admitted by frozen slice.

The full receipt is byte-identical on CPython 3.11.15 and 3.14.8, with optimized workers and a minimal relocated checkout containing spaces. CLI source-pin, missing-log, escaping-symlink and output-preservation controls pass. The 25 existing selected-record tests, three replay-coverage tests and selected-certificate check also pass. Exact source and implementation hashes, commands, finite coverage and the initial temporary-path failure/fix are recorded in the package.

Seed 86, verification engineering. Base 2921874911956718fcaaaf9396fe8267b3cfa9f0; refreshed main abdc6a7002abdbcd9e21bfc5e1a795784deacfdf retains all 12 consumed/context pins byte-for-byte. The final overlap refresh checked 739 open PRs through #1072 and recent coordination comments. This extends the coverage work in #865 to the current T37 log inventory; #991 concerns a different witness/payload boundary and #1006 concerns owner-audit performance.

Credit is retained for Jonathan Capalbo, Chafik Boukhalfa, Hao Yan, Jeremy Kothe and the U8/controller lineage, Douglas Colkitt, Lucas @ljste, Stephen Shore and Adem Vessell. Prepared for Sagar Thakrar with substantial OpenAI Codex assistance. Apache-2.0; predecessor notices are unchanged. Full supplier reconstruction, invoice/outer replay, full make verify, Linux CI and a formal multiplication proof were not performed. All inherited mathematical interfaces remain assumptions.

[Mister-Meeseeks]

100-PR batch review at 89ddfca5207452ce26939d968f512f22c9ef1a12.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: T37 execution-index/log coverage supplement and frozen native-predicate omission controls; not a full-binder or mathematical counterexample.

The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions.

Batch review, attribution and evidence. Thank you for the contribution.