diff --git a/ROADMAP.md b/ROADMAP.md
deleted file mode 100644
index d3a512f..0000000
--- a/ROADMAP.md
+++ /dev/null
@@ -1,47 +0,0 @@
-# Roadmap / Ideas — devcontainer
-
-Backlog of future work — not an active plan. Items here are known gaps or open questions,
-not yet scoped or scheduled.
-
-- [ ] **release.yml: tag pushes still get rejected when their ref history touches a workflow
-  file.** The actual GitHub restriction here needs a token with the classic PAT `workflow`
-  OAuth scope, or a GitHub App installation token (via `actions/create-github-app-token`) —
-  there's no `permissions:` YAML key for this on the default `GITHUB_TOKEN` (confirmed against
-  GitHub's own workflow schema; a `workflows: write` permission block doesn't exist and just
-  makes the whole file invalid — hit that for real, reverted it). Needs a secret provisioned
-  before this can be fixed properly.
-
-- [ ] **AGENTS.md's "verify version after merge" is still a manual step.** PR#52's root cause
-  (a bump landing back at the old value after merge) was never pinned down, so the only
-  safeguard right now is a human remembering to diff `origin/main` after every merge.
-  `release.yml`'s `detect` job already re-diffs every `devcontainer-feature.json` version on
-  push to `main` for its own purposes — extending it to flag "this looks like a bump that
-  should have landed didn't" would close the gap for real.
-
-- [ ] **`peon-ping`'s `host.docker.internal` fix pushes a `runArgs` line onto every consumer**
-  instead of patching `/etc/hosts` automatically at container start. Five other features in
-  this repo already use `postCreateCommand`/`postStartCommand` for exactly this "must act on
-  the live container" problem — worth checking whether the same pattern works here before
-  asking every consumer to edit their own `devcontainer.json`.
-
-- [ ] **AGENTS.md's feature table hand-maintains a `Ver` column** that duplicates each
-  feature's own `devcontainer-feature.json` version, with nothing checking the two stay in
-  sync. Low stakes on its own, but it's the same kind of duplication that let PR#52 drift
-  silently — either generate the column or have the `shellcheck` CI job verify it.
-
-- [ ] **`nub`: dedicated feature vs. folded into an existing one.** Asked for, not delivered
-  yet — a real analysis of whether a standalone `nub` feature earns its keep versus, say,
-  extending `package-auto-install` or `typescript-dev` to cover the same ground.
-
-- [ ] **`pnpm-store` compatibility with `nub`.** Does `nub` delegating to `pnpm` respect the
-  `store-dir` config `pnpm-store` sets up, or does each invocation end up with its own?
-  Don't document the pairing as supported until someone's actually run it.
-
-- [ ] **OpenSSF Scorecard.** Not set up anywhere in the org yet — `helpers4/typescript` has
-  the workflow file but nothing registered on OpenSSF's side, so there's no working example
-  to copy. Wait until that's sorted somewhere first.
-
-- [ ] **`peon-ping` host relay reachability.** Whether the relay is actually reachable (not
-  just DNS-resolvable) depends on what interface `peon relay --daemon` binds to on the host.
-  Needs a real speaker and a real host to test — see the README's "Testing the audio path"
-  section.