docs(adr): ADR 0160 records the public-repo content policy that was being enforced unrecorded

[wshallwshall]

Records a policy that has been in force and enforced while being written down nowhere.

Why this exists

You stated it twice on 2026-08-04: the public repo carries only what someone running MessageFoundry needs plus what a security review needs, with ADRs named explicitly as the second category, and material about how the project is built with Claude Code comes out.

It was recorded nowhere -- a session grepped every local and remote ref and found nothing -- while I was citing it as a constraint on four other sessions' work. That is exactly the standing I had refused from a peer earlier the same day, when I declined to arm a PR on a relayed owner ruling. I applied the rule to others and not to myself, and to an isolated session an unrecorded policy is indistinguishable from an invented one. That asymmetry is the defect this fixes.

The question is new because the 2026-07-27 cutover retired the publish-time deny-list that used to answer it. It is now answered per file, at authoring time, with no gate and no written rule.

Status is Proposed on purpose

The ruling (D1) is yours and already in force. The application (D2/D3) -- which files, in what order -- is mine, and no file moves before you ratify it.

Measured, not recalled

At c90dcb5f: docs/ 580 tracked, docs/adr/ 156 (stays), docs/releases/ 101 -- not the 96 I had been carrying -- scripts/ 61.

Two findings that changed the ADR

The Phase 1 citation surface is 23 files, not the one I first named. Citations split across two grep forms and neither alone is sufficient:

Form	Files
docs/releases/...	12
relative (releases/...	15 (invisible to the above)
union	23

My own verification of my own claim used the single form, returned nothing, and nearly made me delete a true statement as unfounded -- docs/FEATURE-MAP.md:5,:13 do cite it, via the relative form. docs/README.md is in the union, so the docs front door breaks without the fix.

Phase 1 cannot be docs-only, structurally. tests/test_cutover_slug_rot.py:53 carries "docs/releases/" as a literal scan-scope entry, plus a second test and a harness profile. So the PR necessarily touches code -- which satisfies the separate rule that a doc migration must not be docs-only, but the ADR says explicitly not to rely on that, since the guards would run because of a test file this phase happens to touch, not any property of the phase.

D3 records a conflict rather than papering over it

By the policy CLAUDE.md should come out. It cannot: git worktree add cannot deliver an untracked file, so while it was untracked every new worktree came up with zero project conventions loaded -- verified on two live worktrees, and it only looked right in the primary because an untracked copy happened to sit on disk. It stays tracked as a named exception with a recorded mechanical cause, because an untracked convention file is a control that silently fails to load.

D4 proposes no enforcement gate, and says why

A path gate is evaded or wrong; a keyword gate fires on every legitimate use of "session". That is a bounded claim about scope, not a promise of a later mechanism.

Blocking question for you

Does docs/BACKLOG.md stay? It is process material by the test and CI-load-bearing -- backlog_status_check.py plus backlog-hygiene.yml, a required merge context -- so removing it retires a gate rather than editing a doc. This blocks Phase 2.

Three further questions are listed under To resolve on acceptance, including whether docs/Secure_Development_Standards.md is reviewer material that stays (it was deliberately pulled from the public PyPI sdist, which suggests the question was asked once and answered differently).

⚠️ Explicitly not a confidentiality control -- git log retains everything removed, nothing here was ever treated as secret, and anything genuinely sensitive belongs in the vault and always did.