docs(readme): list the five routes, their evidence, and what the VM tier proves

[TusanHomichi]

What this adds

README only. No route, fix, box, adapter or tool changes.

Route: none (documentation)
New fixes: none
Claims issue: none. Refs #13 (the README now summarizes its VM-tier results)

The README still described two routes (12 and 25 fixes) and said kernels, bootloaders and reboots were untested. This brings it up to date:

• Routes: a table of the five routes with fix counts, boxes and run evidence. Every box passes its route's current content id from scratch. I checked each run's route against tools/check.py --ids at f7278c9, and each counted run covers every hop. CI-attested runs: both Debian boxes and centos5-web. The rest are local runs from before the attestation rule, and the README says so.
• Tiers: CI's chroot tier proves userland only; the maintainer-run VM tier covers kernels, bootloaders and real reboots. The README lists the boxes with a from-scratch VM pass on record, at the route versions recorded there: etch-web, etch-db, trusty-db, trusty-web-updated, centos5-web, centos7-lamp and centos8-lamp on BIOS (Routes don't test kernels, bootloaders or a real first boot #13, route: el centos8 -> alma10 #92), plus all three Ubuntu boxes on UEFI (hardware: Ubuntu routes boot on UEFI as on BIOS #84).
• Run provenance: the intro and the runs/ layout line no longer say every run came from CI; new runs do.
• Stale claim removed: CentOS 5 services no longer stay down after a reboot. service.record and service.enable carry SysV runlevels onto systemd (SPEC open question 10).
• Layout: now matches the tree. It adds the Ubuntu adapter, adapters/drivers and adapters/leapp, archive/, docs/rebuild-and-carry.md, tools/vm.py, and the other tools. The client line now says it walks a root filesystem but not yet the running system.

Model

none: documentation change, no fixes and no discovery runs.

Evidence

• python3 wend check at the head, under umask 022 (what CI runners have): exit 0, ending 68 fixes, 5 routes, 10 boxes, 84 runs: ok. An earlier version of this description said the gate's own tests passed, based on that last line. That was wrong for this host's default umask 0002: there it exits 1 because tools/test_net_ifcfg.py depends on umask, though it still prints ok. Filed as wend check can end with 'ok' and still exit 1; test_net_ifcfg depends on umask #128; the same happens on main.
• python3 tools/affected.py origin/main: {"include": []}, so a README change replays nothing.
• ./wend verify was not run: it needs root for the sandbox before it reads the (empty) matrix above.

Self-review (AGENTS.md step 5)

• Every precondition is as narrow as I can make it, and ./wend verify found no decoy hits (n/a: no fixes)
• Every fix fixes the real problem, not the symptom a check inspects (n/a: no fixes)
• One problem per fix; load-bearing steps (like repair) kept (n/a: no fixes)
• No edits to tools/, adapters/, SPEC.md, .github/, other boxes' roles, or other people's fixes (except fires_on)
• Commits are signed off (git commit -s)

Review

An independent read-only review of 9b706a0 found that the CentOS 8 VM pass (#92) was missing, that the VM results were stated per route instead of per box, and that the intro and layout still said every run came from CI. 1820bd1 and 3a10ef6 fix those. The stale SPEC bullet on Ubuntu's VM tier is filed as #129, since SPEC.md is owner-reserved.

🤖 Generated with Claude Code

[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 2026-10-16T05:40:00.494984Z	80fbbe0	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.

[TusanHomichi]

Claude review: SIGN-OFF on head 8f2f13b

An independent read-only review (a separate reviewer agent, not the author) checked this PR against the repo and the GitHub record. It ran in three passes: 9b706a0, then 3a10ef6, then the delta to 8f2f13b.

• Route table: the fix counts match each routes/*.toml. Every listed box has a passing run of its route's current content id (tools/check.py --ids) that covers every hop with model_calls 0. The CI and local labels match runner.kind. The local runs predate the attestation rule (ab1747b).
• VM tier: the box lists are exactly the from-scratch passes on record: BIOS in Routes don't test kernels, bootloaders or a real first boot #13 and route: el centos8 -> alma10 #92, UEFI in hardware: Ubuntu routes boot on UEFI as on BIOS #84. trusty-web's BIOS run was chained resumes, so it is left out of the BIOS list.
• Fixes after the first pass: the first pass found the CentOS 8 VM pass missing, the VM results stated per route, and wording saying every run came from CI. All are fixed. The removed CentOS 5 reboot claim is stale per SPEC's service.record/service.enable.
• Layout block: every path exists and every description matches its module docstring and wend's usage.
• Checks: at 8f2f13b, (umask 022; python3 wend check) exits 0 and tools/affected.py origin/main prints {"include": []}. All commits are signed off.

Filed separately: #128 (wend check prints ok but exits 1 under umask 0002) and #129 (SPEC's stale Ubuntu "container tier only" bullet).

No blocking findings.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8f2f13bd59

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".

[TusanHomichi]

Claude review: SIGN-OFF on head 80fbbe0

The independent reviewer checked the one commit since its sign-off on 8f2f13b. It deletes "#13 stays open until the VM tier runs in CI", answering Codex's P2 (the thread is answered and resolved). SPEC.md "VM tier" (lines 1615-1616 on main) says the tier is maintainer-run only and never wired to a CI runner a pull request can trigger.

With whitespace collapsed, the clause is the only change. No other sentence implies CI VM runs. The commit is signed off. (umask 022; python3 wend check) exits 0, and the PR still replays nothing.

No blocking findings. Every earlier point in the sign-off on 8f2f13b still holds.

🤖 Generated with Claude Code

[TusanHomichi]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Keep it up!

Reviewed commit: 80fbbe0ca6

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".