docs: composability boundary + public-clone deployment anchor

- vision: 'ALPN as a service' section (core is front-door-blind; v1
  reduces to storage + ACL + protocol adapters); primary deployment
  target = public anonymous fetch, authenticated push, minimal ACL
- alk-stack: concrete composability boundary (core+transport depend on
  alkcall types only; adapters replaceable)
- AGENTS.md: convention 15 codifies the boundary

Verified: none needed (markdown only)

docs/research/vision.md

@@ -45,7 +45,8 @@ is designed from day one, not retrofitted.
 3. **No plaintext secrets at rest** — alkvault or nothing.
 4. **Thin interfaces, one core** — http and ssh are adapters that authenticate,
    resolve a repo, and hand a duplex stream to the transport layer. Policy
-   lives in exactly one place.
+   lives in exactly one place. The core never knows which front door is
+   talking (see "ALPN as a service" below).
 5. **Honest capability advertisement** — the git protocol advertises only
    what we actually serve (this is both protocol correctness and the
    security pattern: promise/enforce in the same place).
@@ -75,6 +76,41 @@ is designed from day one, not retrofitted.
 | `alkgit-ssh` | ssh front door (git commands over alkcall channels) |
 | `alkgitd` | binary: config, assembly, TLS/ACME, serving loops |
 
+## Composability: "ALPN as a service"
+
+alkgit is the git member of the alk "ALPN as a service" family — alktty,
+alktunnels, and alksocks (socks5) follow the same pattern, and the alknet
+mono-repo is being decomposed into exactly these pieces for rewrite. The
+pattern's load-bearing rule: **`alkgit-core` + `alkgit-transport` never know
+which front door is talking.** The transport layer consumes (identity, repo
+id, duplex stream, limits) and speaks git; everything above the stream is
+the adapter's problem. That is what keeps alkgit composable for downstream
+use:
+
+- A future gitea/gitlab-like application should be able to embed
+  `alkgit-core` + `alkgit-transport` (or talk to `alkgitd`) and add its own
+  UI/issues/PR layer without forking anything here.
+- The admin API is a set of alkcall ops, not an embedded web framework —
+  a downstream app can either use it or replace it.
+- Nothing in the core may reach upward into alkhttp/alkssh concerns
+  (no http types, no channel types below the transport boundary).
+
+The v1 reduction follows from this: **storage + ACL + protocol adapters**.
+A simple static/template web UI for the public repo listing is explicitly
+out of scope here (would be a separate downstream thing on top).
+
+## Primary deployment target
+
+Our own use case is the design anchor: **public self-hosted repos with
+anonymous clone over http, authenticated push, no gitea-style app features
+in use**. So:
+
+- Anonymous *fetch* (clone/fetch advertisement + pack) on explicitly-public
+  repos is a first-class path, not an afterthought.
+- Push is always authenticated, on every repo, no exceptions.
+- ACL must be simple enough to reason about completely: repo visibility
+  (public/private) + identity-based read/write, nothing richer in v1.
+
 ## What phase 0 must still produce before phase 1
 
 - [x] gitoxide capability + version research (`gitoxide.md`)