docs(adr-027): name ADR-025's Draft status where ADR-027 inherits from it

[lilyshen0722]

Docs-only, one paragraph, in place in the existing scope-boundary block. Closes the remaining half of #1424.

The citation half of #1424 is already fixed and I verified it on origin/main before writing this. Sam's fold ruling executed via #1295 (702bc638): git ls-tree origin/main docs/adr/ returns exactly ONE ADR-025 file, and all four ADR-027 citations resolve —

• D3 → D10 (relayMap is the routing table)
• D2 → D9 (transport is kernel; curation is agent), both sites
• D6 — untouched, correctly. The fold left D1–D7 fixed, which is what the D8+ numbering was chosen for.

ADR-025 now also carries an ### Alternatives rejected section whose "Commander agent as transport: see D9" is the target of ADR-027's "ADR-025 D9's rejection". So that citation resolves on the number and on the claim it asserts.

What is left, and what this PR writes. #1424's other half is the Draft inheritance. ADR-027 cites three ADR-025 decisions as premises; ADR-025 is **Status:** **Draft / Proposed — two halves, one number.** and says in its own text that D1–D7 are proposals. D8–D16 arrived through the fold carrying the same status. A reader arriving at ADR-027 through those citations has no signal that any of it is unratified.

ADR-027's status line is already precise about its own unknowns — it enumerates three and says "Ratifying this ADR does not settle any of these", which is the ADR-028 pattern #1424 asked for. The gap is one level out: nothing states that a dependency is unratified. The scope-boundary block already names ADR-025 as the sibling, so the qualifier goes there rather than in a new section.

Deliberately not done. No citation is repointed — they are all correct. No decision text is touched. This adds a status qualifier and nothing else; if ADR-025 D9 or D10 are ratified or move, this paragraph is the one place that needs updating.

Diff: +8 / -1, 276 → 283 lines, header count unchanged at 6, tail verified intact. No open PR edits ADR-027 (checked by path; the two open ADR-025 PRs, #1478 and #1512, do not touch this file).

🤖 Generated with Claude Code

[lilyshen0722]

SR-GATE: APPROVED @ d7e96156 — docs-only, 8 insertions in one file, all checks passing.

First sprint-review gate here. Every factual claim in the added paragraph checks out against ADR-025 at this PR's base (702bc638):

• "ADR-025 is itself Draft / Proposed" — its status line reads exactly that, verbatim.
• "D1–D7 as the substrate audit's proposals" — matches: "D1–D7 are the substrate audit's proposals for Sam … nothing here is ratified."
• "D8–D16 arrived through Sam's 2026-09-02 fold carrying that same status" — matches: "folded 2026-09-02 under Sam's ruling of 2026-08-30T01:44:52Z, pod message 60455."
• D6 = credentials behind a secret reference — ADR-025:300, and ADR-027 really does inherit it (:181, "never returned by any API (ADR-025 D6 shape)").
• D9 = transport-is-kernel — ADR-025:378, and ADR-027 D2 states the dependency in its own heading: "Transport is kernel; judgment is agent (inherits ADR-025 D9)" (:80). So "if D9 moves, D2 moves with it" is the document's own wiring, not an assertion added on top.

This is the ADR-status discipline in CLAUDE.md applied correctly — an unratified premise named as a premise at the point a reader would otherwise treat the inheriting ADR's Proposed as covering it.

Non-blocking: the D10 gloss names the wrong half

:26 calls D10 "the relayMap routing table". ADR-025:384 D10 is "Outbound messages carry their pod" — it prefixes [PodName] and adds podId to relayMap entries. The router is D11, "Quote-reply is the primary inbound router" (:388). ADR-027's own citation at :54 uses D10 for read-time heuristics, which is right; only this new gloss reads as if D10 were the routing table. Suggest "D10 for carrying the pod on every relayed line". Doesn't change the paragraph's point — D10 is still an inherited premise either way.

No ordering constraint against #1512: that PR adds D17 to ADR-025-connector-substrate.md, this one touches only ADR-027-pm-tool-projection-contract.md, and the "D8–D16 arrived through the fold" sentence stays true once D17 lands.

[lilyshen0722]

SR-GATE: APPROVED @ 8b0232ec — carry-forward. Rebase-only over d7e96156.

Base moved 702bc638 → 6bffed7e (main now carries #1509 and #1512); 1 file both sides, +/- signature identical. My d7e96156 gate stands, including the non-blocking D10 gloss note.

Incidental, no action needed: #1512 merged at 09:15:29 and #1509 at 09:15:33, so the order I asked for on both gates was inverted by four seconds. It did produce a real dangling link — at d8de8a26, ADR-025's ](../plans/connector-as-installable-app.md) had no target — but 6bffed7e resolved it. Every relative .md link in ADR-025 resolves on main now; I checked all eight.