📝 docs(e10): GitHub addressing & representation model (E10-S00) - #85

Parking this PR — it is blocked on a repo-wide CI break, not on this diff.

verify (required) fails because .github/workflows/verify.yaml:56 uses go-version: stable while line 27 pins GOLANGCI_LINT_VERSION: v2.12.2. stable rolled Go 1.26 → 1.27.0 since the last green main run (5ba9da0, 2026-08-19), and golangci-lint 2.12.2 predates Go 1.27 — its bundled typechecker cannot parse the 1.27 stdlib:

/opt/hostedtoolcache/go/1.27.0/x64/src/crypto/internal/randutil/randutil.go:11:2:
  could not import math/rand/v2
  (/opt/hostedtoolcache/go/1.27.0/x64/src/math/rand/v2/rand.go:213:17:
   method must have no type parameters) (typecheck)

That path is inside the Go toolchain’s own source tree. This PR changes zero Go bytes (six Markdown files), so it cannot be the cause.

Evidence:

• The other three required checks — Analyze (go), Analyze (actions), CodeQL — all pass.
• Full local gate matrix is green: task check 19/19 stages exit 0, task changelog-verify ok, task lint reports 0 issues under Go 1.26.6.
• Control: running the pinned golangci-lint 2.12.2 against a Go 1.27.0 GOROOT locally, over a Go tree byte-identical to main’s, panics in pkg/goanalysis/runner_loadingpackage.go:335 — same root cause, different symptom.

The fix (bump the golangci-lint pin, and/or replace go-version: stable with an explicit pin) is deliberately not made here: it would break this lane’s docs/spec-only fence and touch pins guarded by hack/lint/workflow_pins_test.sh. It needs its own change and its own review.

Leaving this open — it should merge unchanged once the toolchain fix lands on main.

Review record

This PR's review happened outside GitHub, so reviewDecision is empty and it reads as unreviewed. It is not. Recording the history here so the maintainer-LGTM decision can be made on the evidence.

Two independent review rounds, a different reviewer each round.

Round 1 — REQUEST CHANGES. One P1: a capability was graded full on the strength of predicates that did not actually decide the property, which contradicted the project's own dossier verdict of partial. Four required P2s alongside it.

Round 2 — APPROVE, no P0/P1. The lane fixed all five findings and chose the fail-closed remedy for the P1: the capability is now graded unknown, so v1 GitHub comments and does not gate. A fresh reviewer verified that the demotion is load-bearing rather than cosmetic, independently re-derived every citation anchor in the document, and confirmed that the replacement conformance case is genuinely falsifiable.

Follow-ups recorded, none blocking: R-1 through R-4, plus R-8, are carried into E10-S01 / E10-S06.

Check state

The last verify run on this head predates the CI toolchain fix (D-158, golangci-lint v2.13.1 for Go 1.27) and failed with that signature; the other required checks passed. A branch update from main was attempted via the update-branch API and was refused with HTTP 422, merge conflict between base and head — the branch needs a real conflict resolution, which has deliberately not been done here. Treat the current check results as stale.

This PR carries a maintainer-LGTM governance marker (openspec/specs/backlog.md:557 — S00/S02/S04 require maintainer LGTM and must be surfaced rather than auto-merged) and is intentionally awaiting the maintainer. It has not been auto-merged.

Maintainer LGTM — granted, and recorded where it is visible

openspec/specs/backlog.md:557 marks E10-S00 (GOVERNANCE) maintainer LGTM. The maintainer
(@konih) granted that LGTM explicitly and authorized this integration.

GitHub's reviewDecision on this PR is empty and will stay empty — the maintainer is also the PR
author, so GitHub refuses a self-review. That is precisely why this comment and D-161 exist:
an approval that exists only in an agent transcript is invisible to the person who later asks
whether the marker was honoured. The durable record is D-161 in docs/decisions/decisions.md.

How it was brought up to date — merge-down, not rebase

The head was 2 ahead / 24 behind with three genuine conflicts. Force-push is forbidden
(AGENTS.md hard rule 2), so instead of a rebase or a replacement PR, main was merged down into
the lane (30c14ee). This keeps #85 — and with it the review history posted above, which a
replacement PR would have stranded.

All three conflicts were pure append-adjacency and resolved as ordered unions, then verified
by content, not by count (this repo has eaten decision rows before and rerere.enabled is false):

• docs/decisions/decisions.md — D-152…D-160 each present exactly once, D-155 sorted into numeric order
• docs/planning/open-questions.md — OQ-31…OQ-36 each exactly once (no collision: this lane minted 33/34, main minted 35/36)
• CHANGELOG.md — all 11 entries present (2 from this lane, 9 from main)
• openspec/specs/backlog.md auto-merged; the E10-S00 DONE row and the GOVERNANCE marker both survive

task changelog-write was then a byte no-op, independently corroborating the CHANGELOG union.
task check is green (exit 0) on the final tree.

The earlier red verify was the stale pre-D-158 toolchain signature, not a real failure; it clears
with the merge-down.

Scope of this approval

Covers E10-S00 only — docs/spec-only, zero Go, zero schema bytes. Its one review P1 was fixed
fail-closed, so v1 GitHub comments and does not gate. Revert is one git revert.

It is not a blanket waiver: the same marker still guards E10-S02/S04, E11-S02/S04/S06/S07,
DEM-S00, WG-S01. Each needs its own LGTM.