[abruption]

Summary

Closes #298 when merged. Prepare an unpublished v1.1.2 reliability patch against published v1.1.1; no merge or publication is performed by this PR.

• Share bounded Claude PID parsing across registry reads, discovery, diagnostics, numeric targeting and the experimental SSH handoff's existing Claude profile checks.
• Accept decimal PID lexemes up to 10 characters, including Unicode decimal digits and bounded leading zeros; reject bool, invalid or out-of-range native PIDs before OS inspection (POSIX signed 32-bit / Windows DWORD).
• Treat nondecimal Unicode names such as ² as names, retain healthy sessions beside abnormal entries, and preserve permission diagnostics and Windows process-start/PID-reuse checks.
• Regenerate standalone source, set candidate package version 1.1.2, and prepare release notes/runbooks in English, Korean, Japanese and Simplified Chinese. README retains published 1.1.1 until publication is verified.

Type

• Bug fix
• Tests and release preparation documentation

Verification

• The exact published v1.1.1 source reproduces the new Unicode filename/name and invalid PID failures in owned offline fixtures; no native message was posted.
• Python 3.13.7: Claude records 16 tests; handoff remote 32 tests; docs 61 tests; distribution 41 tests (3 skipped), all successful.
• Python 3.9.6: core suite 559 tests successful (3 skipped).
• Generated standalone check and git diff --check pass.
• Built wheel and sdist, then installed each in a separate clean environment and checked installed imports with python -I, version 1.1.2 and parser behavior.
• Final head a1b483f5a2028a1c2ea00267a0d21f506d546355: Python 3.13.7 full suite with Relay/MCP and pinned Paramiko test dependencies: 1,104 tests successful (18 skipped); pip check passes.
• Initial full-suite attempt found outdated release-command expectations and a missing test-only packaging dependency. Expectations were updated, dependencies installed in an isolated environment, and the final complete suite passed. No product test failure is waived.
• Exact-head GitHub checks: 28/28 SUCCESS, including platform CI and the aggregate release gate. Automatic code/security review remains pending (no published findings at this observation). GitHub automatically triggers Codex review; no extra mention is sent.

Boundaries and release gates

These are offline tests, not new native-agent, SSH-host, Windows live delivery, Relay/fleet or operational acceptance. Windows native bounds are covered by simulated unit checks pending platform CI. No Codex transport, receipt/ACK contract, permission policy or resend eligibility is expanded; uncertain outcomes never authorize automatic resend.

Do not merge, create a tag/release, publish to PyPI or deploy services without separate authorization. Require normal repository protections, final-head review/CI and exact-main release gates, immutable asset/provenance checks, current dependency audits and required human PyPI environment review. Trusted Publisher configuration and live attestations have not been reverified by this preparation work. Existing experimental PRs and other worktrees are out of scope.

[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 2026-10-13T06:30:27.427018Z	a1b483f	PR opened
🔒 Security Review	✅ Completed 2026-10-13T06:29:05.260991Z	a1b483f	PR opened

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]

v1.1.2 publication verified

• Exact merged/tagged source: f3993bf8988dd301b7e14402e63e8ac04e5e2300.
• GitHub v1.1.2: stable, Latest, immutable; seven assets verified against source, hashes, provenance and authenticated build attestations before publication.
• Main CI, preparation and PyPI publication/post-publication verification: completed/success.
• The required pypi human gate was directly observed waiting at 2026-10-13T06:40:47Z; API approval record identifies reviewer abruption. No gate bypass was used. Current Trusted Publisher UI recheck was explicitly waived by the owner.
• PyPI 1.1.2 exposes exactly the expected, non-yanked wheel and sdist; latest is 1.1.2. Wheel uploaded 2026-10-13T06:41:32.122145Z; sdist 06:41:34.226915Z.
• Downloaded public bytes match locked release hashes: wheel b282f6cdcfbab116a632f4b63178baf2599bb0dd51299d7e77494f1db81ddd66; sdist 8d6ff8ba18f2d66f6b1edc9decc686fa5210e53e9d71b82b06b024e39214a135.
• Both public artifacts were installed independently in clean local environments. Isolated installed imports (python -I), session-peer --version, parser regression checks, initialized empty-home JSON listing and dependency compatibility checks passed. Independent read-only metadata review also confirmed publication/source/hash agreement.

The brief initial PyPI 404 was resolved by propagation; the upload was not repeated. No live native agent message, server/Control/Relay deployment, fleet update, global install or general handoff acceptance is claimed. Tagged README/preparation notes retain their pre-publication wording; the release/PyPI records above establish actual publication.