[abruption]

Summary

Synchronize the runtime repository's compatibility skill copy with the separately published Python skill v0.5.0, so it documents the now-published runtime 1.4.0 MCP workflows and structured preview.

• Copy canonical session-peer/SKILL.md byte-for-byte into skills/session-peer/SKILL.md.
• Update the existing exact-snapshot metadata regression to the published size, SHA-256 and skill version. Keep the bounded frontmatter reader and its limits unchanged.
• Metadata is skill 0.5.0, runtime minimum 0.9.1, documented full coverage 1.4.0; receiver codexBin floor remains 1.0.1.

No runtime code/version, independently installed skill, service, TS file or existing published runtime asset is changed. Historical release notes retain the compatibility snapshot used at their release. This PR does not authorize another runtime or skill publication.

Published source provenance

• Companion release: https://github.com/abruption/session-peer-skill/releases/tag/v0.5.0 (stable, published 2026-11-02T04:57:32Z).
• Annotated tag object: fda9c6421c24eb8a4f58f6125e5f703c92252e1e.
• Peeled source: ee2d16e954b6a0204f9a085ad3c308f4fba83ccd.
• Canonical file: https://github.com/abruption/session-peer-skill/blob/ee2d16e954b6a0204f9a085ad3c308f4fba83ccd/session-peer/SKILL.md
• Git blob: 7ad20c9ae0cf3b874c849c2c8c5a2fc08e602188; UTF-8 bytes: 29326.
• SHA-256: 532d370c3acf18f59d2b53213bfa9b08a877fcb3415de2f3fb607219c665ca0d.

The source release API reports immutable=false. This is not represented as GitHub-enforced immutability: both tag-ref and exact-commit contents were independently downloaded and compared with the local copy, including the full blob/digest above. No repository protection setting was changed. The canonical text is kept exact, including its existing adoption wording; the hosting flag and that wording remain distinct review considerations before merge.

Supporting runtime is published v1.4.0, source 476c67e636ef02cb93797dd2f42252dfe577625f. Existing v1.4.0 GitHub/PyPI artifacts retain the old compatibility copy; no asset is replaced here.

Validation

Python 3.13.7, isolated temporary paths:

• python -m unittest tests.core.test_skill_frontmatter tests.distribution.test_install tests.distribution.test_verified_release tests.distribution.test_release_artifacts -q: 42 tests pass.
• python -m unittest tests.core.test_skill_frontmatter -v: 15 tests pass, including the updated published snapshot test.
• python -m unittest discover -s tests/docs -q: 89 tests pass.
• python tools/generate_session_peer.py --check; git diff --check: pass.
• Canonical bytes via tag and pinned commit both compare identically. Existing stale pins outside this test are dated release-preparation records, not current snapshot guards; plugins contain README files, no second canonical skill copy.

The initial targeted run correctly failed the old v0.4.1 exact-size assertion; after updating that existing pin to the independently verified v0.5.0 data, the same suite passed. No expectations were weakened.

Independent review

Following the current independent-cli-review skill, Claude Code 2.1.314 / claude-opus-5-5 and Antigravity CLI 1.4.12 / gemini-4-argon-medium, medium, ran non-interactively in parallel over the same frozen 73054-byte packet. Input SHA-256: 32cf2174344fe83004cd670359e0de1d4f69e437fff8e0a0bb5c05a8ebb87550, unchanged after both runs.

Claude tools/MCP/hooks were disabled; Antigravity used fresh plan+sandbox with a 600s limit. Both exited 0 with successful substantive responses; the requested models were confirmed in Claude modelUsage and Antigravity init.model. Elapsed times: Claude 106.544s, Antigravity 46.221s. No external-action tool execution was observed.

Neither found a required source edit. Codex checked Claude's provenance, old-reference, anchor and command-evidence questions against the repository/API, including the immutable=false qualification above. Two non-blocking upstream wording suggestions concern making per-tool permissions and optional preflight clearer; they are not local edits to this exact published snapshot. Antigravity's textual packaging inference is not a claim that a future archive hash is unchanged: tests can be included in an sdist, and published assets are preserved independently of future builds.

Review is source/document analysis, not new native/MCP/SSH/ACK acceptance, global installation or operational validation. Automatic current-head CI and review remain separate gates. No manual Codex mention is needed.

[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 2026-11-02T05:07:19.944692Z	4aaae35	PR opened
🔒 Security Review	✅ Completed 2026-11-02T05:08:12.096442Z	4aaae35	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.