Tested Rust source: 38789512626cb24f78f274998977fc8c6cf960d1. The plan precedes implementation; the result and reproduction instructions state the acceptance limits. This is a private read-only optical experiment, not an editable workspace or release package.

Final content inventory covers 29 files, 17465556 stored bytes and 1805 archived members / 86189726 uncompressed bytes. The original index and original two archives remain unchanged; a separate supplement adds the later independent checker review and parent observations. Final index SHA-256: 7bf2380b49c307fd24ff0289586668359e24947f40630f9f05442bc74a6205a1. The index excludes its own recursive hash; this README and the portable verifier are outside the raw packet inventory.

The later category-stroke critique is also outside that frozen packet. It examines a possible next hypothesis against the exact frozen scene and primary Capella/Sirius sources; no dash implementation or perception acceptance is claimed.

The handoff narrative review pins its exact inspected document hashes and finds no narrative blocker. The first publication check verifies unchanged source inputs and an empty successful staged whitespace check; later navigation-pointer additions do not claim retrospective review. Both are outside the frozen raw index.

Archive / record	Contents
raw-native.zip	Five actual native case directories, 62 original screenshots, complete optical inputs/models/scenes/SVGs/measurements, watcher/Close receipts, logs, gate and formatting records, final mechanical oracle
oracle-challenges.zip	Original and repaired checker's isolated synthetic challenge corpus, exact mutations, failures and result records; these are not native runs
supplemental-final-review.zip	Complete independent repair-review probe corpus, original report/probe, parent associations and their write helper
source review	Independent preformat source handoff review; original scope retained
evaluator review and repair	Original defects, authorized repair and fourteen copied-corpus falsifiers
independent final checker review	Four actual cases, 22 captures, 533 gate-source hashes and twelve further falsifiers; explicitly limited checker/evidence scope
parent observations	Assisted native notes and exact SHA-bound associations for all 22 measurement/screenshot pairs, including unsuccessful probes

Executables/runtime are local-only, excluded from the packet, and pinned in provenance. Host-specific original helpers and paths are preserved; relocating them requires an explicit adaptation. A future build is a new result, not automatically the historically pinned executable.

Verify content without extraction or executable launch, from any host with Python3:

python harness/experiments/focus-optical-20261001/verify_evidence.py

On the original evidence host, --originals also compares all 1823 retained archived/copied source files. After staging the packet, --git-index additionally compares its exact staged bytes. Synthetic adversarial data must never replace the native corpus when running the actual mechanical checker. Byte integrity verification makes no new semantic, rendering or readability claim.