# Independent package readback

2026-09-25. PASS: no actionable package discrepancy found. Reviewed source `8d52dab97c92cc31baae9f20d8c324b55818aa5d` and the immutable evidence ZIP: **65 members, 479,039 bytes**, SHA256 `8d46b31379d75567b195b639cb3bc289ad1c9d26df53d29dbfd99ba22e38c5b9`. Manifest SHA256: `ac6df312f15efee4c1656d01ddb26379d653d88593edf003f8617447947a476c`.

Independently verified ZIP CRC, unique safe member names, exact manifest membership, and every member length/hash. All 63 file members match their original local gate, baseline, diagnostic, contract, review or packaging-helper bytes. Both remaining members match exact Git diffs (`fa7a980418ef99d772e67c66c28dfa7b5181f875` to final; `19a4082117b88caf332198e4c685bde91005cb52` to final). All seven readable mirrors match both archive and original review files. Baseline records also match the previously published carried-phase ZIP, SHA256 `2eca143b706458c1c0217af715fa6f17c01a0b9f85a5313d6a231dabe1aba7c9`.

Independent in-memory readback verifies all **555 declared inputs** against immutable Git blobs and current raw files, final HEAD and clean tracked source, receipt/log/input/launcher hashes, and final gate status. Raw results contain **253 passed / 16 failed**, exactly **11 new focused passes**, and **269 listed names**. All 258 previous outcomes and all 16 failure blocks match the baseline, normalizing only numeric thread identifiers. Strict core all-target Clippy exits 0. Formatting exits 1 with byte-identical baseline output, SHA256 `c1c6b12861c8117ce6ffc6d6742ebf92618aafb10160ec6f99d370ca72ebab2d`. This remains a failed full suite and failed formatting check.

The initial full run remains an abort: 135 completed passes followed by exit `3221225725` (`0xc00000fd`). Its separately retained 42,402,816-byte artifact hashes to `777457cfefb8a3046e3f0809cabd936cfb742abe792befad2c50e05dfac1bf6b`; the equally sized final artifact hashes to `29d8a4761fd8f09a3c0eafeb11cca07f0d45c711bf1b969d97c8b29d2b3ab446`. Both identities were checked against actual bytes without execution. The debugger trace retains Exact decimal parsing and binding frames plus explicit process termination; debugger exit zero is not a passing test.

Archived disassembly binds initial binding's 27,272-byte reservation and final binding/prefix reservations of 2,520 + 11,720 = 14,240 bytes: a 13,032-byte reduction in those selected frames. Continuation is mutually exclusive with prefix and reserves 13,416 bytes. These figures are not total live stack use. Each of the three separate control receipts/logs records the exact unchanged ordinary-index test passing in a fresh process, with final source/artifact identity and matching helper hash. The inspected helper rejects stack overrides and stops on failure; it was not executed during this audit.

README, relevant STATE and DECISIONS entries preserve private **unverified** output scope, initial failure, inherited failures, the separate older unrerun setup limitation, and pending actual-output parsing/resolution/equality/ownership obligations. No public editing, universal stack safety, broad before-growth accounting, Linux execution or new release is inferred. Relative Markdown links in the packet reviews/README and root STATE/DECISIONS resolve with this readback present. Historical `PID` labels in the runtime checker/report refer to the normalized Rust thread identifier and do not change the comparison.

Only read-only Git, source/data inspection, hashing and independent Python assertions were used. No archived helper, Cargo, test/product executable, debugger or network operation was run; the existing runtime checker's output was not overwritten. The ZIP, manifest, gate records, code and original reviews remain unchanged. Publication remains the parent's responsibility.