[abruption]

Summary

Curates the isolated macOS observations from 2026-10-13 for #307. Adds an archive README and three unchanged probe snapshots under experiments/antigravity-startup/, excluded from wheel and sdist. No runtime, hook configuration, permission, i18n contract or version change.

The record distinguishes:

• No externally observable conversation ID/hook at the recorded pre-input idle checkpoints (not proof that no internal ID exists).
• SessionStart registration after first input, without proof of native-send authority.
• A receiver-tool-runner-started bridge with an on-screen owner ACK on pinned Antigravity 1.3.4.
• A different TUI's direct agentapi invocation parented by the caller, with no target-owner receipt observed during the bounded test.

Refs #307; the issue remains open. This does not establish pre-input delivery, general startup support or stable Antigravity support. #304 / PR #305 remain separate.

Evidence and limits

• Historical runtime: published v1.1.1 (b48144bdf2ad1569031562136f5a60e33780020e); archive base: main f3993bf8988dd301b7e14402e63e8ac04e5e2300.
• The initial TUI displayed 1.3.3 while the installed executable reported 1.3.4 afterward; that first send is not an exact-version result. Follow-ups used a pinned 1.3.4 copy.
• Reported/reconstructed results, on-screen ACKs and body-free metadata are explicitly separated. Raw native/send stdout is not published. consumptionConfirmed remained false.
• Source custody distinguishes a reconstructed initial collaborator probe from retained joint-run source. Snapshot equality does not establish identity with every historical execution. Observer revisions are disclosed.
• Private evidence/manifests remain private. Disposable test resources were reported cleaned; this PR does not claim all evidence was deleted or independently verify historical settings restoration.
• Archived probes are unhardened snapshots, not an approved future-run recipe: fixed markers, unbounded native output capture and file/readiness race limitations are disclosed. No probe or native agent was executed for this change.
• Linux/Windows, later versions, long-term stability, legitimate hook authority and structured receipts remain unestablished. Neither upstream documentation link was freshly fetched for this archive.

Validation

• Three probe snapshots compare byte-identically with the preserved original experiment worktree; existing untracked evidence was not edited.
• Probe compilation only (py_compile), generator --check, git diff --check: PASS.
• Documentation tests: 61 PASS, rerun after final wording edits.
• Isolated wheel/sdist build: PASS. Distribution tests with the built artifacts: 41 tests OK, 3 PowerShell skips. No experiments entries in either artifact. Local artifacts are verification-only and will not be published.
• A first build-environment dependency probe lacked hatchling; the subsequent project-pinned isolated build succeeded. This was not a runtime failure.

Independent review

Used independent-cli-review, with identical frozen stdin per round, no reviewer-authorized writes, tests, network or native submissions. CLI versions: Claude Code 2.1.300; Codex CLI 0.162.1. Claude requested/confirmed first-round model: claude-opus-5-5. Codex used --ignore-user-config --sandbox read-only --ephemeral; its actual default model was not exposed by the result metadata.

• First round, candidate 950b07242333fb277caf4d2d0ad266301fdced30, input SHA-256 04cb1b61f2ba5e559e936f6795f367831743bae566800a6ebaf6c5858bedfa4a: both completed, exit 0, valid final reviews; no mandatory finding. UTC starts 01:48:37, Codex ended 01:49:03, Claude ended 01:49:48 on 2026-10-15.
• Follow-up, candidate 8d246f0032b11ed81bc3d70b020496e646ceef45, input SHA-256 5ac5dde97e3abc439bdf62d5c7738a31a2c163fb6b367c104be4ad15b322ea00: Codex completed with no mandatory finding (01:51:21–01:51:43 UTC). Claude exited 0 but attempted an unavailable tool and returned cancellation text rather than a valid review (ended 01:52:33 UTC); that follow-up is not counted as a completed review, and its model usage also included claude-opus-4-8. The valid first-round Opus 5.5 review is not replaced by it.
• Root checked the recommendations against the files and clarified receipt/timing bounds, private metadata, historical references and context-variable semantics. Final small wording edits were locally checked, not independently re-reviewed by both CLIs. These are static reviews, not new authentication, native-receipt or platform tests.

Current-head CI and automatic review must be assessed separately. No manual review mention, merge, release, tag, installation, publication or operational change is included.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-16T00:33:06.886247Z	f7e7a0e	Manual request
🔒 Security Review	✅ Completed 2026-10-16T00:33:31.609801Z	f7e7a0e	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[abruption]

@codex review

[abruption]

@codex security review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Swish!

Reviewed commit: f7e7a0eefd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

🛡️ Codex Security Review

Security review completed. No security issues were found in this pull request.

Reviewed commit: f7e7a0eefd

View security finding report

Only the user who started this review can view the report in Codex.

ℹ️ About Codex security reviews in GitHub

This is an experimental Codex feature. Security reviews are triggered when:

• You comment "@codex security review"
• A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review

Once complete, Codex will leave suggestions, or a comment if no findings are found.