[JacobStephens2]

Summary

- Native Apple Silicon macOS binary and LaunchAgent scheduling
+ Linux musl binaries: x86_64 and aarch64
+ Apple Silicon: Ubuntu 24.04 aarch64 in Lima (vz, --mount-none)
+ CI test suite: Ubuntu; release smoke checks only: Rocky Linux and WSL2

Align ADR 0003, README, manual and site platform guidance. Keep repositories on the Linux filesystem, remove native macOS caveats and scheduling instructions, and regenerate the HTML manual. Update the hand-written architecture/how-it-works pages and agent index too.

Closes #1668

Evidence

• Before: The homepage structured-data test failed with actual Linux, macOS, expected Linux.
  After: The same test passes with operatingSystem: "Linux".
• Before: The Spec review's exact Python assertion failed on both manual files' advice to run the Console on other operating systems. The retained Node regression also failed.
  After: Both checks pass with the Console kept inside Linux VM/WSL2.
• npm run docs:check --prefix site: passed.
• npm test --prefix site on final HEAD: 61 tests passed, including the retained regression.
• The affected-test selection also passed cargo build --locked and all 126 Console browser tests.
• CI run 41110152864 passed on final HEAD 00e8a2d6: the full Rust suite (cargo nextest run), formatting, clippy, Python tooling, site and Console checks.
• scripts/affected-tests main was attempted twice locally and did not finish green. The first run stopped when the host's Python 3.9 could not import tomllib; settlement::historical_review_and_reclosed_origin_regressions passes with Python 3.11. The second run hit session::recording::tests::reader_failure_keeps_models_consumed_before_the_log_failed, which passed in the first run and passes in isolation with --test-threads=3. Per docs/testing.md, the successful CI full suite at this exact commit is reused instead of running it a third time.

Local validation selects Python 3.11 on PATH and uses PLAYWRIGHT_BROWSERS_PATH=/home/jstephens/.cache/ms-playwright on this host.

Test seams

• Homepage JSON-LD as consumed without JavaScript.
• Published manual guidance in Markdown and generated HTML.

Review coverage

Independent thirdshift-code-review axes used issue-1665 as the fixed point. Standards: 0 findings. Spec: 1 finding, reproduced and fixed with a retained regression test. Both axes read all 10 changed files, including generated site/docs/index.html. Generated-output exclusions: none. Unread changed files: none.

Security review used thirdshift-security-audit in guidance mode, in one session, against pinned merge base 86c8887f557b225573a32552bb3fb0927bc04edf. It examined all 10 changed files, including the generated manual, and supporting documentation generation, Console binding/installation, child executable selection, process identity, session cleanup and sandbox-probe code. No introduced Security findings were identified. git diff --check 86c8887f557b225573a32552bb3fb0927bc04edf...HEAD passed. Validation reuses the successful CI run linked above at the exact reviewed HEAD 00e8a2d6, as permitted by docs/testing.md.

Unaddressed findings

• Standards: None.
• Spec: None.
• Security: None.

Security review outcome

No unaddressed introduced findings.

Merge Danger

Door: two-way

Documentation and test changes can be reverted.

Blast Radius: documentation

Changes installation, platform coverage and scheduling guidance across the public docs.

Built with codex codex-cli 0.166.0 · gpt-6.1-sol · xhigh