# Local verification record — 2026-09-26

> Historical pre-publication evidence. For the 2026-10-03 release checks and
> external publication results, see [release record](RELEASE_0.1.0.md).

Environment: Windows, `moon 0.1.20260827`, `moonc v0.10.11+6ff76a5f9`.
These are local observations, not a remote CI result or security audit.

| Check | Local result |
| --- | --- |
| `moon fmt --check` | Pass |
| `moon check` | Pass |
| `moon test --deny-warn` | 109/109 pass |
| `moon check --target all --deny-warn` | Pass (Wasm, Wasm-GC, JS, Native) |
| `moon build --target wasm-gc --deny-warn` | Pass |
| `moon build --target native --release --deny-warn` | Pass |
| `moon test --target wasm-gc --deny-warn` | 109/109 pass |
| `moon test --target js --deny-warn` | 109/109 pass |
| `moon test --target native --release --deny-warn` | 109/109 pass |
| `moon test client/refresh_test.mbt --target wasm --deny-warn` | 10/10 pass; targeted only |
| `moon run examples/offline` | Pass; five single-target cases and one bundle |
| `moon run --target js examples/offline` | Pass; same scenarios |
| `moon run --target native --release examples/offline` | Pass; same scenarios |
| `moon info` | Pass; generated `*.generated.mbti` files are ignored |

The non-test MoonBit count is **4,533 effective lines**: include executable
examples, exclude `*_test.mbt`, `testkit/`, blank lines and `//`/`///` comment
lines. This is a reproducible local scale measure, not a claim that line count
alone proves quality. The local Git history has more than 20 substantive
development commits; run `git rev-list --count HEAD` for the current count.

This run includes an independent RFC 8032 Ed25519 known-answer vector,
regressions for online-key rotation, root-only rotation, and trusted-state
gating while a root chain is incomplete. Additional regressions cover resumed
refresh after missing snapshot/targets, same-version pin stability, public
JSON getter isolation, delegation DFS/termination/budgets, and repository
names returned by refresh for rotated roots and consistent snapshots. Three
additional tests check that unfinished root chains and expired timestamp or
snapshot parents cannot advance their child metadata stage.
Parser and trust regressions also cover large signature lists, late duplicate
signers, undeclared role keys, duplicate delegated roles and many unknown
signatures preceding one valid authorized signer.
Two refresh regressions distinguish an authoritative absent successor root
from a failed fetch, both before and after one authenticated rotation. The
loader API and security guidance now require failures to stop root probing.

The Native Debug example required noticeably more time than Native Release
on this machine. Neither run is a controlled benchmark; see
[`PERFORMANCE.md`](PERFORMANCE.md) for complexity and measurement limits.

Intentionally incomplete external acceptance items: no GitHub/Gitlink remote
or public default-branch check, no remote CI run, no Mooncakes publication,
and no independent TUF conformance suite/security audit. The upstream suite's
CLI/storage/HTTP requirements and fixture limitations are documented in
[`research/tuf-conformance-adapter.md`](../research/tuf-conformance-adapter.md),
not reported as a passing run. The proposal also
keeps the applicant identity and future public links as explicit fields to
finalize before submission. These items are **not** represented as passing.