Branch-specific evidence for the implementation report. Executed Scenario tests belong to 67534ce4; final lint, format, Studio/CLI build and native runs belong to 8d6fac47. The report explains the intervening test annotations and UI text changes. This is not combined integration or release acceptance.

Archive	Bytes	SHA-256
`checks/evidence.zip`	12897498	`d3de224502a2b7f8236d2ad82af6d6c57bb3a349468aebeea7369c3c3c267342`
`native/native-evidence.zip`	6069072	`27e0e189053fdfdbfcecd8c91ae718849448b090aab557afc7d9445af1ceaba8`

Run from a checkout with Python 3.11+; these checks need no Cargo, Git, network or installed Studio:

python -B harness/scenario-group-spacing-evidence-20260924/checks/verify.py
python -B harness/scenario-group-spacing-evidence-20260924/native/verify.py harness/scenario-group-spacing-evidence-20260924/native

Checks contain ten closed runs, two preserved failures, ten exact source proofs and four deduplicated source capsules. Explicit run names are group-spacing-format-1, group-spacing-scenario-1, group-spacing-format-2, group-spacing-scenario-2, group-spacing-clippy-1, group-spacing-format-3, group-spacing-clippy-2, group-spacing-format-4, group-spacing-build-1 and group-spacing-clippy-3. The reused checks helper README's selection example/title and manifest kind are historical schema names; the actual run inventory and report govern this package. Failed-run consistency is not a passing test result.

Native evidence retains both normal-exit processes, all 67 captures, actual saved snapshots, immutable expected fixtures, current and obsolete-CLI preflights, receipts and original helpers. The helpers' archived preparation notes saying “not run” describe their original preparation stage; the subsequent closed receipts establish execution. Executable/DLL bytes are excluded explicitly, with exact pins and source/build bindings retained. This is a portable evidence archive, not a runnable distribution. The offline checker independently compares actual saved source with declared phases and validates fresh-reopen provenance; it does not judge visual usability or rehash excluded binaries.

The independent review distinguishes source review, native observations and package consistency. Reproduction/environment and visual/input limitations are in the implementation report. Raw screenshot labels are identifiers, not assertions (run-1 capture35 clears the group to one visible member, not zero selection).