Skip to content

docs: the M12 record — a tailnet login that lasts until the token is rotated, and the roadmap row (#230) - #237

Merged
davison merged 3 commits into
mainfrom
docs/m12-record
Sep 30, 2026
Merged

davison merged 3 commits into
mainfrom
docs/m12-record

Conversation

@davison

@davison davison commented Sep 30, 2026 •

Copy link
Copy Markdown
Owner

The M12 record, on the housekeeping path (SPEC §4). Milestone: #230. This PR does not close it; gh codecrew milestone close 12 does.

What this adds

Gaps the record names

The coordinator's attribution Decision on #230 (where the operator's words came from) exists, and the record cites it for each attribution. It was posted late, at record time, and the record says so under Where the record is silent.

Checks

  • gh codecrew milestone evidence 12: requirements M12-R1 to R5 counted, all 6 cited links on the trail resolve. The record is not on main yet, so each of its 35 distinct github.com links was also resolved with gh api, and each comment link checked to belong to the issue or PR it names.
  • make check exits 0.

🤖 Generated with Claude Code

davison and others added 2 commits September 30, 2026 12:46
…rotated, and the roadmap row (#230)

This is the milestone record on the housekeeping path (SPEC section 4):
docs/milestones/12-a-tailnet-login-that-lasts.md, the M12 row in
ROADMAP.md, and the front-door claims M12 changed. The introduction's
opening describes the project at the end of milestone twelve, with a
paragraph on what M12 changed, and its suite count goes from 120 to 123;
CONTRIBUTING lists openssl for the tailnet session suite and counts
twelve suites and 123 tests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The milestone record on the housekeeping path. The coordinator's
Decision on where the operator's words came from (#230, 5910571213)
now backs the gate resolution, the manual link and the wait for it,
the relayed scope ("ov v0.3.1", quoted corrected), the upstream-capture
ask and the tool-written operator confirmation. The record keeps that
the attribution was posted late, at record time.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@davison

davison commented Sep 30, 2026

Copy link
Copy Markdown
Owner Author

Review — 1d8a237

Reviewer seat (~, operator auth), clean context, in a detached worktree at 1d8a237. This was the head I was given, and it is still the PR head. gh codecrew version reports protocol 2.1, which matches the floor in .codecrew/config.yml. I read the diff first. Then I read the charter: #230's body and all 5 comments (scope, QA verdicts, observations, disposition, attribution). Then #231 (plan, both Decisions, gate, resolution, Deviation, hand-close, QA finding, full timeline), PRs #232, #233 and #234 (bodies, all comments, both review rounds, #232's timeline), captures #229, #235 and #236, upstream radiusred/gh-codecrew#386 and #318, and the #39 session Decision. I read the PR body last. I held the record to the doc-synthesizer contract, including this hub's front-door list, and used the M11 record and its review as the bar.

Findings

1. nit: the PR body contradicts the record it carries. Its Gaps the record names section says "Unlike M11, no attribution comment is on the trail for the operator's session words. The record gives the attributions as the coordinator relayed them to this seat and says so". That was true of e5a7664. 1d8a237 rewrote the record to cite the attribution Decision (#230), and Where the record is silent now says that Decision exists and was posted late. The Checks section also says "each of its 34 distinct github.com links". The record at head has 35 (grep -oE 'https://github\.com/[^) >]+' … | sort -u | wc -l → 35), because the Decision link was added. Fix: edit the PR body. Say the attribution Decision exists and was posted at record time, and give the count as 35. This does not touch the branch.

2. nit: an unwrapped line in the introduction. docs/introduction.md:1414 is 107 columns ("running or stopped; that suite also needs openssl, for its certificate. Its own harness is tested too: a"). The rest of What holds these numbers wraps at 87 or less on main (awk over lines 1340–1420). This is the same kind of nit round one on #232 raised for docs/running.md. Fix: reflow the paragraph.

3. nit: the manual-link wait is stated more strongly than its source. docs/milestones/12-a-tailnet-login-that-lasts.md:282-284 says "From about 10:44Z to 11:23Z nothing else happened on #231". The attribution Decision says only that "the task waited on that until the connected event". #231's timeline has one more event in that window: referenced 2026-09-30T10:44:58Z cf8d5159…, the reflow commit pushed after round two's approval. The record describes that commit itself at :64-68. Fix: "the task waited on that link until the connected event", with no "nothing else happened". Or start the window at about 10:45Z.

4. nit: two claims in The hand-close cite a comment that does not carry them. :290-294 says task finish reported that the task "closes via its closing keyword" without checking, and that "the coordinator noticed by reading the issue's state". The only citation is the hand-close comment (#231). That comment says neither thing. Both come from radiusred/gh-codecrew#386 ("prints merged PR #N; <task> closes via its closing keyword" and "The coordinator only noticed by reading the issue state afterwards"). Fix: cite #386 in that paragraph as well.

5. nit: "the Decision corrected it" is ambiguous. At :39-41, "He typed "ov v0.3.1", and the Decision corrected it to "of"", with the citation going to the attribution Decision. It was the scope Decision (5909079844) that quoted it with "of". The attribution Decision only records that it did. Also, "corrected" is the record's own word: the source says "quotes it with "of"". Fix: "He typed "ov v0.3.1"; the scope Decision quotes it with "of" ([attribution Decision])", which matches the wording the record already uses at :443-444.

6. nit: one of the #39 quotes is not verbatim. At :141, "Nothing to revoke short of rotating." is given as #39's objection. #39 reads "nothing to revoke either short of rotating" (#39). The record took the #231 Decision's shortened form. The other three quoted objections are verbatim. Fix: restore "either", or drop the quotation marks.

7. nit: the sync guide is misquoted in the front-door section. :390-392 says "The e-ink and sync guides say a session is a cookie that survives closing the tab". Only docs/e-ink.md:104 says "closing the tab". docs/sync.md:323-324 says "The session is a cookie and survives what follows", meaning the app-install steps. Both still read true after M12, so the conclusion stands. Fix: quote each guide on its own terms.

Nothing blocking.

What I verified, and how

Capture candidates

None.

Verdict: approve

The milestone record on the housekeeping path. The introduction's suite
paragraph is reflowed. The record says the task waited on the manual
link, with cf8d515 the one other event in that window; cites
radiusred/gh-codecrew#386 for the "closes via its closing keyword"
output and how the hand-close was noticed; says the scope Decision
quotes "ov" with "of"; restores "either" in the #39 quote; and quotes
the e-ink and sync guides each on its own terms.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@davison

davison commented Sep 30, 2026

Copy link
Copy Markdown
Owner Author

Operator confirmation: this record, at head 9a38e03, is accepted for the operator (pure solo tier, SPEC §4). The acceptance rests on his standing merge confirmation for this milestone and its record ("run it through to a release"), recorded in #230 (comment) and relayed from the coordinator session, per #230 (comment).

The model review approved at 1d8a237 (#237 (comment)). Commit 9a38e03 takes all seven of that review's nits, which are wording and line wrapping only. The gate passed; check and e2e were skipped because every commit is docs. The doc-synthesizer may merge.

@davison
davison merged commit 8385614 into main Sep 30, 2026
3 checks passed
@davison
davison deleted the docs/m12-record branch September 30, 2026 11:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Tailnet sessions that survive restarts and end on rotation

1 participant